Synchronization and Inter-Layer Interactions of Noise-Driven Neural Networks.

PubMed

Yuniati, Anis; Mai, Te-Lun; Chen, Chi-Ming

2017-01-01

In this study, we used the Hodgkin-Huxley (HH) model of neurons to investigate the phase diagram of a developing single-layer neural network and that of a network consisting of two weakly coupled neural layers. These networks are noise driven and learn through the spike-timing-dependent plasticity (STDP) or the inverse STDP rules. We described how these networks transited from a non-synchronous background activity state (BAS) to a synchronous firing state (SFS) by varying the network connectivity and the learning efficacy. In particular, we studied the interaction between a SFS layer and a BAS layer, and investigated how synchronous firing dynamics was induced in the BAS layer. We further investigated the effect of the inter-layer interaction on a BAS to SFS repair mechanism by considering three types of neuron positioning (random, grid, and lognormal distributions) and two types of inter-layer connections (random and preferential connections). Among these scenarios, we concluded that the repair mechanism has the largest effect for a network with the lognormal neuron positioning and the preferential inter-layer connections.

A one-layer recurrent neural network for constrained pseudoconvex optimization and its application for dynamic portfolio optimization.

PubMed

Liu, Qingshan; Guo, Zhishan; Wang, Jun

2012-02-01

In this paper, a one-layer recurrent neural network is proposed for solving pseudoconvex optimization problems subject to linear equality and bound constraints. Compared with the existing neural networks for optimization (e.g., the projection neural networks), the proposed neural network is capable of solving more general pseudoconvex optimization problems with equality and bound constraints. Moreover, it is capable of solving constrained fractional programming problems as a special case. The convergence of the state variables of the proposed neural network to achieve solution optimality is guaranteed as long as the designed parameters in the model are larger than the derived lower bounds. Numerical examples with simulation results illustrate the effectiveness and characteristics of the proposed neural network. In addition, an application for dynamic portfolio optimization is discussed. Copyright © 2011 Elsevier Ltd. All rights reserved.

Target recognition based on convolutional neural network

NASA Astrophysics Data System (ADS)

Wang, Liqiang; Wang, Xin; Xi, Fubiao; Dong, Jian

2017-11-01

One of the important part of object target recognition is the feature extraction, which can be classified into feature extraction and automatic feature extraction. The traditional neural network is one of the automatic feature extraction methods, while it causes high possibility of over-fitting due to the global connection. The deep learning algorithm used in this paper is a hierarchical automatic feature extraction method, trained with the layer-by-layer convolutional neural network (CNN), which can extract the features from lower layers to higher layers. The features are more discriminative and it is beneficial to the object target recognition.

A neural network approach to cloud classification

NASA Technical Reports Server (NTRS)

Lee, Jonathan; Weger, Ronald C.; Sengupta, Sailes K.; Welch, Ronald M.

1990-01-01

It is shown that, using high-spatial-resolution data, very high cloud classification accuracies can be obtained with a neural network approach. A texture-based neural network classifier using only single-channel visible Landsat MSS imagery achieves an overall cloud identification accuracy of 93 percent. Cirrus can be distinguished from boundary layer cloudiness with an accuracy of 96 percent, without the use of an infrared channel. Stratocumulus is retrieved with an accuracy of 92 percent, cumulus at 90 percent. The use of the neural network does not improve cirrus classification accuracy. Rather, its main effect is in the improved separation between stratocumulus and cumulus cloudiness. While most cloud classification algorithms rely on linear parametric schemes, the present study is based on a nonlinear, nonparametric four-layer neural network approach. A three-layer neural network architecture, the nonparametric K-nearest neighbor approach, and the linear stepwise discriminant analysis procedure are compared. A significant finding is that significantly higher accuracies are attained with the nonparametric approaches using only 20 percent of the database as training data, compared to 67 percent of the database in the linear approach.

A novel and generalized approach in the inversion of geoelectrical resistivity data using Artificial Neural Networks (ANN)

NASA Astrophysics Data System (ADS)

Raj, A. Stanley; Srinivas, Y.; Oliver, D. Hudson; Muthuraj, D.

2014-03-01

The non-linear apparent resistivity problem in the subsurface study of the earth takes into account the model parameters in terms of resistivity and thickness of individual subsurface layers using the trained synthetic data by means of Artificial Neural Networks (ANN). Here we used a single layer feed-forward neural network with fast back propagation learning algorithm. So on proper training of back propagation networks it tends to give the resistivity and thickness of the subsurface layer model of the field resistivity data with reference to the synthetic data trained in the appropriate network. During training, the weights and biases of the network are iteratively adjusted to make network performance function level more efficient. On adequate training, errors are minimized and the best result is obtained using the artificial neural networks. The network is trained with more number of VES data and this trained network is demonstrated by the field data. The accuracy of inversion depends upon the number of data trained. In this novel and specially designed algorithm, the interpretation of the vertical electrical sounding has been done successfully with the more accurate layer model.

Gas Classification Using Deep Convolutional Neural Networks.

PubMed

Peng, Pai; Zhao, Xiaojin; Pan, Xiaofang; Ye, Wenbin

2018-01-08

In this work, we propose a novel Deep Convolutional Neural Network (DCNN) tailored for gas classification. Inspired by the great success of DCNN in the field of computer vision, we designed a DCNN with up to 38 layers. In general, the proposed gas neural network, named GasNet, consists of: six convolutional blocks, each block consist of six layers; a pooling layer; and a fully-connected layer. Together, these various layers make up a powerful deep model for gas classification. Experimental results show that the proposed DCNN method is an effective technique for classifying electronic nose data. We also demonstrate that the DCNN method can provide higher classification accuracy than comparable Support Vector Machine (SVM) methods and Multiple Layer Perceptron (MLP).

Gas Classification Using Deep Convolutional Neural Networks

PubMed Central

Peng, Pai; Zhao, Xiaojin; Pan, Xiaofang; Ye, Wenbin

2018-01-01

In this work, we propose a novel Deep Convolutional Neural Network (DCNN) tailored for gas classification. Inspired by the great success of DCNN in the field of computer vision, we designed a DCNN with up to 38 layers. In general, the proposed gas neural network, named GasNet, consists of: six convolutional blocks, each block consist of six layers; a pooling layer; and a fully-connected layer. Together, these various layers make up a powerful deep model for gas classification. Experimental results show that the proposed DCNN method is an effective technique for classifying electronic nose data. We also demonstrate that the DCNN method can provide higher classification accuracy than comparable Support Vector Machine (SVM) methods and Multiple Layer Perceptron (MLP). PMID:29316723

Classification capacity of a modular neural network implementing neurally inspired architecture and training rules.

PubMed

Poirazi, Panayiota; Neocleous, Costas; Pattichis, Costantinos S; Schizas, Christos N

2004-05-01

A three-layer neural network (NN) with novel adaptive architecture has been developed. The hidden layer of the network consists of slabs of single neuron models, where neurons within a slab--but not between slabs--have the same type of activation function. The network activation functions in all three layers have adaptable parameters. The network was trained using a biologically inspired, guided-annealing learning rule on a variety of medical data. Good training/testing classification performance was obtained on all data sets tested. The performance achieved was comparable to that of SVM classifiers. It was shown that the adaptive network architecture, inspired from the modular organization often encountered in the mammalian cerebral cortex, can benefit classification performance.

Classification of urine sediment based on convolution neural network

NASA Astrophysics Data System (ADS)

Pan, Jingjing; Jiang, Cunbo; Zhu, Tiantian

2018-04-01

By designing a new convolution neural network framework, this paper breaks the constraints of the original convolution neural network framework requiring large training samples and samples of the same size. Move and cropping the input images, generate the same size of the sub-graph. And then, the generated sub-graph uses the method of dropout, increasing the diversity of samples and preventing the fitting generation. Randomly select some proper subset in the sub-graphic set and ensure that the number of elements in the proper subset is same and the proper subset is not the same. The proper subsets are used as input layers for the convolution neural network. Through the convolution layer, the pooling, the full connection layer and output layer, we can obtained the classification loss rate of test set and training set. In the red blood cells, white blood cells, calcium oxalate crystallization classification experiment, the classification accuracy rate of 97% or more.

Assessing the effect of quantitative and qualitative predictors on gastric cancer individuals survival using hierarchical artificial neural network models.

PubMed

Amiri, Zohreh; Mohammad, Kazem; Mahmoudi, Mahmood; Parsaeian, Mahbubeh; Zeraati, Hojjat

2013-01-01

There are numerous unanswered questions in the application of artificial neural network models for analysis of survival data. In most studies, independent variables have been studied as qualitative dichotomous variables, and results of using discrete and continuous quantitative, ordinal, or multinomial categorical predictive variables in these models are not well understood in comparison to conventional models. This study was designed and conducted to examine the application of these models in order to determine the survival of gastric cancer patients, in comparison to the Cox proportional hazards model. We studied the postoperative survival of 330 gastric cancer patients who suffered surgery at a surgical unit of the Iran Cancer Institute over a five-year period. Covariates of age, gender, history of substance abuse, cancer site, type of pathology, presence of metastasis, stage, and number of complementary treatments were entered in the models, and survival probabilities were calculated at 6, 12, 18, 24, 36, 48, and 60 months using the Cox proportional hazards and neural network models. We estimated coefficients of the Cox model and the weights in the neural network (with 3, 5, and 7 nodes in the hidden layer) in the training group, and used them to derive predictions in the study group. Predictions with these two methods were compared with those of the Kaplan-Meier product limit estimator as the gold standard. Comparisons were performed with the Friedman and Kruskal-Wallis tests. Survival probabilities at different times were determined using the Cox proportional hazards and a neural network with three nodes in the hidden layer; the ratios of standard errors with these two methods to the Kaplan-Meier method were 1.1593 and 1.0071, respectively, revealed a significant difference between Cox and Kaplan-Meier (P < 0.05) and no significant difference between Cox and the neural network, and the neural network and the standard (Kaplan-Meier), as well as better accuracy for the neural network (with 3 nodes in the hidden layer). Probabilities of survival were calculated using three neural network models with 3, 5, and 7 nodes in the hidden layer, and it has been observed that none of the predictions was significantly different from results with the Kaplan-Meier method and they appeared more comparable towards the last months (fifth year). However, we observed better accuracy using the neural network with 5 nodes in the hidden layer. Using the Cox proportional hazards and a neural network with 3 nodes in the hidden layer, we found enhanced accuracy with the neural network model. Neural networks can provide more accurate predictions for survival probabilities compared to the Cox proportional hazards mode, especially now that advances in computer sciences have eliminated limitations associated with complex computations. It is not recommended in order to adding too many hidden layer nodes because sample size related effects can reduce the accuracy. We recommend increasing the number of nodes to a point that increased accuracy continues (decrease in mean standard error), however increasing nodes should cease when a change in this trend is observed.

Growth kinetics of borided layers: Artificial neural network and least square approaches

NASA Astrophysics Data System (ADS)

Campos, I.; Islas, M.; Ramírez, G.; VillaVelázquez, C.; Mota, C.

2007-05-01

The present study evaluates the growth kinetics of the boride layer Fe 2B in AISI 1045 steel, by means of neural networks and the least square techniques. The Fe 2B phase was formed at the material surface using the paste boriding process. The surface boron potential was modified considering different boron paste thicknesses, with exposure times of 2, 4 and 6 h, and treatment temperatures of 1193, 1223 and 1273 K. The neural network and the least square models were set by the layer thickness of Fe 2B phase, and assuming that the growth of the boride layer follows a parabolic law. The reliability of the techniques used is compared with a set of experiments at a temperature of 1223 K with 5 h of treatment time and boron potentials of 2, 3, 4 and 5 mm. The results of the Fe 2B layer thicknesses show a mean error of 5.31% for the neural network and 3.42% for the least square method.

Modeling neural circuits in Parkinson's disease.

PubMed

Psiha, Maria; Vlamos, Panayiotis

2015-01-01

Parkinson's disease (PD) is caused by abnormal neural activity of the basal ganglia which are connected to the cerebral cortex in the brain surface through complex neural circuits. For a better understanding of the pathophysiological mechanisms of PD, it is important to identify the underlying PD neural circuits, and to pinpoint the precise nature of the crucial aberrations in these circuits. In this paper, the general architecture of a hybrid Multilayer Perceptron (MLP) network for modeling the neural circuits in PD is presented. The main idea of the proposed approach is to divide the parkinsonian neural circuitry system into three discrete subsystems: the external stimuli subsystem, the life-threatening events subsystem, and the basal ganglia subsystem. The proposed model, which includes the key roles of brain neural circuit in PD, is based on both feed-back and feed-forward neural networks. Specifically, a three-layer MLP neural network with feedback in the second layer was designed. The feedback in the second layer of this model simulates the dopamine modulatory effect of compacta on striatum.

A One-Layer Recurrent Neural Network for Real-Time Portfolio Optimization With Probability Criterion.

PubMed

Liu, Qingshan; Dang, Chuangyin; Huang, Tingwen

2013-02-01

This paper presents a decision-making model described by a recurrent neural network for dynamic portfolio optimization. The portfolio-optimization problem is first converted into a constrained fractional programming problem. Since the objective function in the programming problem is not convex, the traditional optimization techniques are no longer applicable for solving this problem. Fortunately, the objective function in the fractional programming is pseudoconvex on the feasible region. It leads to a one-layer recurrent neural network modeled by means of a discontinuous dynamic system. To ensure the optimal solutions for portfolio optimization, the convergence of the proposed neural network is analyzed and proved. In fact, the neural network guarantees to get the optimal solutions for portfolio-investment advice if some mild conditions are satisfied. A numerical example with simulation results substantiates the effectiveness and illustrates the characteristics of the proposed neural network.

Non-Intrusive Gaze Tracking Using Artificial Neural Networks

DTIC Science & Technology

1994-01-05

We have developed an artificial neural network based gaze tracking, system which can be customized to individual users. A three layer feed forward...empirical analysis of the performance of a large number of artificial neural network architectures for this task. Suggestions for further explorations...for neurally based gaze trackers are presented, and are related to other similar artificial neural network applications such as autonomous road following.

Melanoma segmentation based on deep learning.

PubMed

Zhang, Xiaoqing

2017-12-01

Malignant melanoma is one of the most deadly forms of skin cancer, which is one of the world's fastest-growing cancers. Early diagnosis and treatment is critical. In this study, a neural network structure is utilized to construct a broad and accurate basis for the diagnosis of skin cancer, thereby reducing screening errors. The technique is able to improve the efficacy for identification of normally indistinguishable lesions (such as pigment spots) versus clinically unknown lesions, and to ultimately improve the diagnostic accuracy. In the field of medical imaging, in general, using neural networks for image segmentation is relatively rare. The existing traditional machine-learning neural network algorithms still cannot completely solve the problem of information loss, nor detect the precise division of the boundary area. We use an improved neural network framework, described herein, to achieve efficacious feature learning, and satisfactory segmentation of melanoma images. The architecture of the network includes multiple convolution layers, dropout layers, softmax layers, multiple filters, and activation functions. The number of data sets can be increased via rotation of the training set. A non-linear activation function (such as ReLU and ELU) is employed to alleviate the problem of gradient disappearance, and RMSprop/Adam are incorporated to optimize the loss algorithm. A batch normalization layer is added between the convolution layer and the activation layer to solve the problem of gradient disappearance and explosion. Experiments, described herein, show that our improved neural network architecture achieves higher accuracy for segmentation of melanoma images as compared with existing processes.

Efficient and self-adaptive in-situ learning in multilayer memristor neural networks.

PubMed

Li, Can; Belkin, Daniel; Li, Yunning; Yan, Peng; Hu, Miao; Ge, Ning; Jiang, Hao; Montgomery, Eric; Lin, Peng; Wang, Zhongrui; Song, Wenhao; Strachan, John Paul; Barnell, Mark; Wu, Qing; Williams, R Stanley; Yang, J Joshua; Xia, Qiangfei

2018-06-19

Memristors with tunable resistance states are emerging building blocks of artificial neural networks. However, in situ learning on a large-scale multiple-layer memristor network has yet to be demonstrated because of challenges in device property engineering and circuit integration. Here we monolithically integrate hafnium oxide-based memristors with a foundry-made transistor array into a multiple-layer neural network. We experimentally demonstrate in situ learning capability and achieve competitive classification accuracy on a standard machine learning dataset, which further confirms that the training algorithm allows the network to adapt to hardware imperfections. Our simulation using the experimental parameters suggests that a larger network would further increase the classification accuracy. The memristor neural network is a promising hardware platform for artificial intelligence with high speed-energy efficiency.

[The Identification of the Origin of Chinese Wolfberry Based on Infrared Spectral Technology and the Artificial Neural Network].

PubMed

Li, Zhong; Liu, Ming-de; Ji, Shou-xiang

2016-03-01

The Fourier Transform Infrared Spectroscopy (FTIR) is established to find the geographic origins of Chinese wolfberry quickly. In the paper, the 45 samples of Chinese wolfberry from different places of Qinghai Province are to be surveyed by FTIR. The original data matrix of FTIR is pretreated with common preprocessing and wavelet transform. Compared with common windows shifting smoothing preprocessing, standard normal variation correction and multiplicative scatter correction, wavelet transform is an effective spectrum data preprocessing method. Before establishing model through the artificial neural networks, the spectra variables are compressed by means of the wavelet transformation so as to enhance the training speed of the artificial neural networks, and at the same time the related parameters of the artificial neural networks model are also discussed in detail. The survey shows even if the infrared spectroscopy data is compressed to 1/8 of its original data, the spectral information and analytical accuracy are not deteriorated. The compressed spectra variables are used for modeling parameters of the backpropagation artificial neural network (BP-ANN) model and the geographic origins of Chinese wolfberry are used for parameters of export. Three layers of neural network model are built to predict the 10 unknown samples by using the MATLAB neural network toolbox design error back propagation network. The number of hidden layer neurons is 5, and the number of output layer neuron is 1. The transfer function of hidden layer is tansig, while the transfer function of output layer is purelin. Network training function is trainl and the learning function of weights and thresholds is learngdm. net. trainParam. epochs=1 000, while net. trainParam. goal = 0.001. The recognition rate of 100% is to be achieved. It can be concluded that the method is quite suitable for the quick discrimination of producing areas of Chinese wolfberry. The infrared spectral analysis technology combined with the artificial neural networks is proved to be a reliable and new method for the identification of the original place of Traditional Chinese Medicine.

Nested neural networks

NASA Technical Reports Server (NTRS)

Baram, Yoram

1988-01-01

Nested neural networks, consisting of small interconnected subnetworks, allow for the storage and retrieval of neural state patterns of different sizes. The subnetworks are naturally categorized by layers of corresponding to spatial frequencies in the pattern field. The storage capacity and the error correction capability of the subnetworks generally increase with the degree of connectivity between layers (the nesting degree). Storage of only few subpatterns in each subnetworks results in a vast storage capacity of patterns and subpatterns in the nested network, maintaining high stability and error correction capability.

Rod-Shaped Neural Units for Aligned 3D Neural Network Connection.

PubMed

Kato-Negishi, Midori; Onoe, Hiroaki; Ito, Akane; Takeuchi, Shoji

2017-08-01

This paper proposes neural tissue units with aligned nerve fibers (called rod-shaped neural units) that connect neural networks with aligned neurons. To make the proposed units, 3D fiber-shaped neural tissues covered with a calcium alginate hydrogel layer are prepared with a microfluidic system and are cut in an accurate and reproducible manner. These units have aligned nerve fibers inside the hydrogel layer and connectable points on both ends. By connecting the units with a poly(dimethylsiloxane) guide, 3D neural tissues can be constructed and maintained for more than two weeks of culture. In addition, neural networks can be formed between the different neural units via synaptic connections. Experimental results indicate that the proposed rod-shaped neural units are effective tools for the construction of spatially complex connections with aligned nerve fibers in vitro. © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim.

Training Data Requirement for a Neural Network to Predict Aerodynamic Coefficients

NASA Technical Reports Server (NTRS)

Korsmeyer, David (Technical Monitor); Rajkumar, T.; Bardina, Jorge

2003-01-01

Basic aerodynamic coefficients are modeled as functions of angle of attack, speed brake deflection angle, Mach number, and side slip angle. Most of the aerodynamic parameters can be well-fitted using polynomial functions. We previously demonstrated that a neural network is a fast, reliable way of predicting aerodynamic coefficients. We encountered few under fitted and/or over fitted results during prediction. The training data for the neural network are derived from wind tunnel test measurements and numerical simulations. The basic questions that arise are: how many training data points are required to produce an efficient neural network prediction, and which type of transfer functions should be used between the input-hidden layer and hidden-output layer. In this paper, a comparative study of the efficiency of neural network prediction based on different transfer functions and training dataset sizes is presented. The results of the neural network prediction reflect the sensitivity of the architecture, transfer functions, and training dataset size.

Artificial neural network modeling and optimization of ultrahigh pressure extraction of green tea polyphenols.

PubMed

Xi, Jun; Xue, Yujing; Xu, Yinxiang; Shen, Yuhong

2013-11-01

In this study, the ultrahigh pressure extraction of green tea polyphenols was modeled and optimized by a three-layer artificial neural network. A feed-forward neural network trained with an error back-propagation algorithm was used to evaluate the effects of pressure, liquid/solid ratio and ethanol concentration on the total phenolic content of green tea extracts. The neural network coupled with genetic algorithms was also used to optimize the conditions needed to obtain the highest yield of tea polyphenols. The obtained optimal architecture of artificial neural network model involved a feed-forward neural network with three input neurons, one hidden layer with eight neurons and one output layer including single neuron. The trained network gave the minimum value in the MSE of 0.03 and the maximum value in the R(2) of 0.9571, which implied a good agreement between the predicted value and the actual value, and confirmed a good generalization of the network. Based on the combination of neural network and genetic algorithms, the optimum extraction conditions for the highest yield of green tea polyphenols were determined as follows: 498.8 MPa for pressure, 20.8 mL/g for liquid/solid ratio and 53.6% for ethanol concentration. The total phenolic content of the actual measurement under the optimum predicated extraction conditions was 582.4 ± 0.63 mg/g DW, which was well matched with the predicted value (597.2mg/g DW). This suggests that the artificial neural network model described in this work is an efficient quantitative tool to predict the extraction efficiency of green tea polyphenols. Crown Copyright © 2013. Published by Elsevier Ltd. All rights reserved.

A simple method to derive bounds on the size and to train multilayer neural networks

NASA Technical Reports Server (NTRS)

Sartori, Michael A.; Antsaklis, Panos J.

1991-01-01

A new derivation is presented for the bounds on the size of a multilayer neural network to exactly implement an arbitrary training set; namely, the training set can be implemented with zero error with two layers and with the number of the hidden-layer neurons equal to no.1 is greater than p - 1. The derivation does not require the separation of the input space by particular hyperplanes, as in previous derivations. The weights for the hidden layer can be chosen almost arbitrarily, and the weights for the output layer can be found by solving no.1 + 1 linear equations. The method presented exactly solves (M), the multilayer neural network training problem, for any arbitrary training set.

The Variance Reaction Time Model

ERIC Educational Resources Information Center

Sikstrom, Sverker

2004-01-01

The variance reaction time model (VRTM) is proposed to account for various recognition data on reaction time, the mirror effect, receiver-operating-characteristic (ROC) curves, etc. The model is based on simple and plausible assumptions within a neural network: VRTM is a two layer neural network where one layer represents items and one layer…

Optical implementation of a feature-based neural network with application to automatic target recognition

NASA Technical Reports Server (NTRS)

Chao, Tien-Hsin; Stoner, William W.

1993-01-01

An optical neural network based on the neocognitron paradigm is introduced. A novel aspect of the architecture design is shift-invariant multichannel Fourier optical correlation within each processing layer. Multilayer processing is achieved by feeding back the ouput of the feature correlator interatively to the input spatial light modulator and by updating the Fourier filters. By training the neural net with characteristic features extracted from the target images, successful pattern recognition with intraclass fault tolerance and interclass discrimination is achieved. A detailed system description is provided. Experimental demonstrations of a two-layer neural network for space-object discrimination is also presented.

Automatic target recognition using a feature-based optical neural network

NASA Technical Reports Server (NTRS)

Chao, Tien-Hsin

1992-01-01

An optical neural network based upon the Neocognitron paradigm (K. Fukushima et al. 1983) is introduced. A novel aspect of the architectural design is shift-invariant multichannel Fourier optical correlation within each processing layer. Multilayer processing is achieved by iteratively feeding back the output of the feature correlator to the input spatial light modulator and updating the Fourier filters. By training the neural net with characteristic features extracted from the target images, successful pattern recognition with intra-class fault tolerance and inter-class discrimination is achieved. A detailed system description is provided. Experimental demonstration of a two-layer neural network for space objects discrimination is also presented.

Modeling polyvinyl chloride Plasma Modification by Neural Networks

NASA Astrophysics Data System (ADS)

Wang, Changquan

2018-03-01

Neural networks model were constructed to analyze the connection between dielectric barrier discharge parameters and surface properties of material. The experiment data were generated from polyvinyl chloride plasma modification by using uniform design. Discharge voltage, discharge gas gap and treatment time were as neural network input layer parameters. The measured values of contact angle were as the output layer parameters. A nonlinear mathematical model of the surface modification for polyvinyl chloride was developed based upon the neural networks. The optimum model parameters were obtained by the simulation evaluation and error analysis. The results of the optimal model show that the predicted value is very close to the actual test value. The prediction model obtained here are useful for discharge plasma surface modification analysis.

Robust hepatic vessel segmentation using multi deep convolution network

NASA Astrophysics Data System (ADS)

Kitrungrotsakul, Titinunt; Han, Xian-Hua; Iwamoto, Yutaro; Foruzan, Amir Hossein; Lin, Lanfen; Chen, Yen-Wei

2017-03-01

Extraction of blood vessels of the organ is a challenging task in the area of medical image processing. It is really difficult to get accurate vessel segmentation results even with manually labeling by human being. The difficulty of vessels segmentation is the complicated structure of blood vessels and its large variations that make them hard to recognize. In this paper, we present deep artificial neural network architecture to automatically segment the hepatic vessels from computed tomography (CT) image. We proposed novel deep neural network (DNN) architecture for vessel segmentation from a medical CT volume, which consists of three deep convolution neural networks to extract features from difference planes of CT data. The three networks have share features at the first convolution layer but will separately learn their own features in the second layer. All three networks will join again at the top layer. To validate effectiveness and efficiency of our proposed method, we conduct experiments on 12 CT volumes which training data are randomly generate from 5 CT volumes and 7 using for test. Our network can yield an average dice coefficient 0.830, while 3D deep convolution neural network can yield around 0.7 and multi-scale can yield only 0.6.

Neural-Network Simulator

NASA Technical Reports Server (NTRS)

Mitchell, Paul H.

1991-01-01

F77NNS (FORTRAN 77 Neural Network Simulator) computer program simulates popular back-error-propagation neural network. Designed to take advantage of vectorization when used on computers having this capability, also used on any computer equipped with ANSI-77 FORTRAN Compiler. Problems involving matching of patterns or mathematical modeling of systems fit class of problems F77NNS designed to solve. Program has restart capability so neural network solved in stages suitable to user's resources and desires. Enables user to customize patterns of connections between layers of network. Size of neural network F77NNS applied to limited only by amount of random-access memory available to user.

A new optimized GA-RBF neural network algorithm.

PubMed

Jia, Weikuan; Zhao, Dean; Shen, Tian; Su, Chunyang; Hu, Chanli; Zhao, Yuyan

2014-01-01

When confronting the complex problems, radial basis function (RBF) neural network has the advantages of adaptive and self-learning ability, but it is difficult to determine the number of hidden layer neurons, and the weights learning ability from hidden layer to the output layer is low; these deficiencies easily lead to decreasing learning ability and recognition precision. Aiming at this problem, we propose a new optimized RBF neural network algorithm based on genetic algorithm (GA-RBF algorithm), which uses genetic algorithm to optimize the weights and structure of RBF neural network; it chooses new ways of hybrid encoding and optimizing simultaneously. Using the binary encoding encodes the number of the hidden layer's neurons and using real encoding encodes the connection weights. Hidden layer neurons number and connection weights are optimized simultaneously in the new algorithm. However, the connection weights optimization is not complete; we need to use least mean square (LMS) algorithm for further leaning, and finally get a new algorithm model. Using two UCI standard data sets to test the new algorithm, the results show that the new algorithm improves the operating efficiency in dealing with complex problems and also improves the recognition precision, which proves that the new algorithm is valid.

Deep Learning and Developmental Learning: Emergence of Fine-to-Coarse Conceptual Categories at Layers of Deep Belief Network.

PubMed

Sadeghi, Zahra

2016-09-01

In this paper, I investigate conceptual categories derived from developmental processing in a deep neural network. The similarity matrices of deep representation at each layer of neural network are computed and compared with their raw representation. While the clusters generated by raw representation stand at the basic level of abstraction, conceptual categories obtained from deep representation shows a bottom-up transition procedure. Results demonstrate a developmental course of learning from specific to general level of abstraction through learned layers of representations in a deep belief network. © The Author(s) 2016.

Neural network approximation of nonlinearity in laser nano-metrology system based on TLMI

NASA Astrophysics Data System (ADS)

Olyaee, Saeed; Hamedi, Samaneh

2011-02-01

In this paper, an approach based on neural network (NN) for nonlinearity modeling in a nano-metrology system using three-longitudinal-mode laser heterodyne interferometer (TLMI) for length and displacement measurements is presented. We model nonlinearity errors that arise from elliptically and non-orthogonally polarized laser beams, rotational error in the alignment of laser head with respect to the polarizing beam splitter, rotational error in the alignment of the mixing polarizer, and unequal transmission coefficients in the polarizing beam splitter. Here we use a neural network algorithm based on the multi-layer perceptron (MLP) network. The simulation results show that multi-layer feed forward perceptron network is successfully applicable to real noisy interferometer signals.

Cardiac Arrhythmia Classification by Multi-Layer Perceptron and Convolution Neural Networks.

PubMed

Savalia, Shalin; Emamian, Vahid

2018-05-04

The electrocardiogram (ECG) plays an imperative role in the medical field, as it records heart signal over time and is used to discover numerous cardiovascular diseases. If a documented ECG signal has a certain irregularity in its predefined features, this is called arrhythmia, the types of which include tachycardia, bradycardia, supraventricular arrhythmias, and ventricular, etc. This has encouraged us to do research that consists of distinguishing between several arrhythmias by using deep neural network algorithms such as multi-layer perceptron (MLP) and convolution neural network (CNN). The TensorFlow library that was established by Google for deep learning and machine learning is used in python to acquire the algorithms proposed here. The ECG databases accessible at PhysioBank.com and kaggle.com were used for training, testing, and validation of the MLP and CNN algorithms. The proposed algorithm consists of four hidden layers with weights, biases in MLP, and four-layer convolution neural networks which map ECG samples to the different classes of arrhythmia. The accuracy of the algorithm surpasses the performance of the current algorithms that have been developed by other cardiologists in both sensitivity and precision.

Processing of chromatic information in a deep convolutional neural network.

PubMed

Flachot, Alban; Gegenfurtner, Karl R

2018-04-01

Deep convolutional neural networks are a class of machine-learning algorithms capable of solving non-trivial tasks, such as object recognition, with human-like performance. Little is known about the exact computations that deep neural networks learn, and to what extent these computations are similar to the ones performed by the primate brain. Here, we investigate how color information is processed in the different layers of the AlexNet deep neural network, originally trained on object classification of over 1.2M images of objects in their natural contexts. We found that the color-responsive units in the first layer of AlexNet learned linear features and were broadly tuned to two directions in color space, analogously to what is known of color responsive cells in the primate thalamus. Moreover, these directions are decorrelated and lead to statistically efficient representations, similar to the cardinal directions of the second-stage color mechanisms in primates. We also found, in analogy to the early stages of the primate visual system, that chromatic and achromatic information were segregated in the early layers of the network. Units in the higher layers of AlexNet exhibit on average a lower responsivity for color than units at earlier stages.

Performance of Deep and Shallow Neural Networks, the Universal Approximation Theorem, Activity Cliffs, and QSAR.

PubMed

Winkler, David A; Le, Tu C

2017-01-01

Neural networks have generated valuable Quantitative Structure-Activity/Property Relationships (QSAR/QSPR) models for a wide variety of small molecules and materials properties. They have grown in sophistication and many of their initial problems have been overcome by modern mathematical techniques. QSAR studies have almost always used so-called "shallow" neural networks in which there is a single hidden layer between the input and output layers. Recently, a new and potentially paradigm-shifting type of neural network based on Deep Learning has appeared. Deep learning methods have generated impressive improvements in image and voice recognition, and are now being applied to QSAR and QSAR modelling. This paper describes the differences in approach between deep and shallow neural networks, compares their abilities to predict the properties of test sets for 15 large drug data sets (the kaggle set), discusses the results in terms of the Universal Approximation theorem for neural networks, and describes how DNN may ameliorate or remove troublesome "activity cliffs" in QSAR data sets. © 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim.

Application of a neural network for reflectance spectrum classification

NASA Astrophysics Data System (ADS)

Yang, Gefei; Gartley, Michael

2017-05-01

Traditional reflectance spectrum classification algorithms are based on comparing spectrum across the electromagnetic spectrum anywhere from the ultra-violet to the thermal infrared regions. These methods analyze reflectance on a pixel by pixel basis. Inspired by high performance that Convolution Neural Networks (CNN) have demonstrated in image classification, we applied a neural network to analyze directional reflectance pattern images. By using the bidirectional reflectance distribution function (BRDF) data, we can reformulate the 4-dimensional into 2 dimensions, namely incident direction × reflected direction × channels. Meanwhile, RIT's micro-DIRSIG model is utilized to simulate additional training samples for improving the robustness of the neural networks training. Unlike traditional classification by using hand-designed feature extraction with a trainable classifier, neural networks create several layers to learn a feature hierarchy from pixels to classifier and all layers are trained jointly. Hence, the our approach of utilizing the angular features are different to traditional methods utilizing spatial features. Although training processing typically has a large computational cost, simple classifiers work well when subsequently using neural network generated features. Currently, most popular neural networks such as VGG, GoogLeNet and AlexNet are trained based on RGB spatial image data. Our approach aims to build a directional reflectance spectrum based neural network to help us to understand from another perspective. At the end of this paper, we compare the difference among several classifiers and analyze the trade-off among neural networks parameters.

Neural Network Model For Fast Learning And Retrieval

NASA Astrophysics Data System (ADS)

Arsenault, Henri H.; Macukow, Bohdan

1989-05-01

An approach to learning in a multilayer neural network is presented. The proposed network learns by creating interconnections between the input layer and the intermediate layer. In one of the new storage prescriptions proposed, interconnections are excitatory (positive) only and the weights depend on the stored patterns. In the intermediate layer each mother cell is responsible for one stored pattern. Mutually interconnected neurons in the intermediate layer perform a winner-take-all operation, taking into account correlations between stored vectors. The performance of networks using this interconnection prescription is compared with two previously proposed schemes, one using inhibitory connections at the output and one using all-or-nothing interconnections. The network can be used as a content-addressable memory or as a symbolic substitution system that yields an arbitrarily defined output for any input. The training of a model to perform Boolean logical operations is also described. Computer simulations using the network as an autoassociative content-addressable memory show the model to be efficient. Content-addressable associative memories and neural logic modules can be combined to perform logic operations on highly corrupted data.

Detecting central fixation by means of artificial neural networks in a pediatric vision screener using retinal birefringence scanning.

PubMed

Gramatikov, Boris I

2017-04-27

Reliable detection of central fixation and eye alignment is essential in the diagnosis of amblyopia ("lazy eye"), which can lead to blindness. Our lab has developed and reported earlier a pediatric vision screener that performs scanning of the retina around the fovea and analyzes changes in the polarization state of light as the scan progresses. Depending on the direction of gaze and the instrument design, the screener produces several signal frequencies that can be utilized in the detection of central fixation. The objective of this study was to compare artificial neural networks with classical statistical methods, with respect to their ability to detect central fixation reliably. A classical feedforward, pattern recognition, two-layer neural network architecture was used, consisting of one hidden layer and one output layer. The network has four inputs, representing normalized spectral powers at four signal frequencies generated during retinal birefringence scanning. The hidden layer contains four neurons. The output suggests presence or absence of central fixation. Backpropagation was used to train the network, using the gradient descent algorithm and the cross-entropy error as the performance function. The network was trained, validated and tested on a set of controlled calibration data obtained from 600 measurements from ten eyes in a previous study, and was additionally tested on a clinical set of 78 eyes, independently diagnosed by an ophthalmologist. In the first part of this study, a neural network was designed around the calibration set. With a proper architecture and training, the network provided performance that was comparable to classical statistical methods, allowing perfect separation between the central and paracentral fixation data, with both the sensitivity and the specificity of the instrument being 100%. In the second part of the study, the neural network was applied to the clinical data. It allowed reliable separation between normal subjects and affected subjects, its accuracy again matching that of the statistical methods. With a proper choice of a neural network architecture and a good, uncontaminated training data set, the artificial neural network can be an efficient classification tool for detecting central fixation based on retinal birefringence scanning.

Egg production forecasting: Determining efficient modeling approaches.

PubMed

Ahmad, H A

2011-12-01

Several mathematical or statistical and artificial intelligence models were developed to compare egg production forecasts in commercial layers. Initial data for these models were collected from a comparative layer trial on commercial strains conducted at the Poultry Research Farms, Auburn University. Simulated data were produced to represent new scenarios by using means and SD of egg production of the 22 commercial strains. From the simulated data, random examples were generated for neural network training and testing for the weekly egg production prediction from wk 22 to 36. Three neural network architectures-back-propagation-3, Ward-5, and the general regression neural network-were compared for their efficiency to forecast egg production, along with other traditional models. The general regression neural network gave the best-fitting line, which almost overlapped with the commercial egg production data, with an R(2) of 0.71. The general regression neural network-predicted curve was compared with original egg production data, the average curves of white-shelled and brown-shelled strains, linear regression predictions, and the Gompertz nonlinear model. The general regression neural network was superior in all these comparisons and may be the model of choice if the initial overprediction is managed efficiently. In general, neural network models are efficient, are easy to use, require fewer data, and are practical under farm management conditions to forecast egg production.

Deep learning in the small sample size setting: cascaded feed forward neural networks for medical image segmentation

NASA Astrophysics Data System (ADS)

Gaonkar, Bilwaj; Hovda, David; Martin, Neil; Macyszyn, Luke

2016-03-01

Deep Learning, refers to large set of neural network based algorithms, have emerged as promising machine- learning tools in the general imaging and computer vision domains. Convolutional neural networks (CNNs), a specific class of deep learning algorithms, have been extremely effective in object recognition and localization in natural images. A characteristic feature of CNNs, is the use of a locally connected multi layer topology that is inspired by the animal visual cortex (the most powerful vision system in existence). While CNNs, perform admirably in object identification and localization tasks, typically require training on extremely large datasets. Unfortunately, in medical image analysis, large datasets are either unavailable or are extremely expensive to obtain. Further, the primary tasks in medical imaging are organ identification and segmentation from 3D scans, which are different from the standard computer vision tasks of object recognition. Thus, in order to translate the advantages of deep learning to medical image analysis, there is a need to develop deep network topologies and training methodologies, that are geared towards medical imaging related tasks and can work in a setting where dataset sizes are relatively small. In this paper, we present a technique for stacked supervised training of deep feed forward neural networks for segmenting organs from medical scans. Each `neural network layer' in the stack is trained to identify a sub region of the original image, that contains the organ of interest. By layering several such stacks together a very deep neural network is constructed. Such a network can be used to identify extremely small regions of interest in extremely large images, inspite of a lack of clear contrast in the signal or easily identifiable shape characteristics. What is even more intriguing is that the network stack achieves accurate segmentation even when it is trained on a single image with manually labelled ground truth. We validate this approach,using a publicly available head and neck CT dataset. We also show that a deep neural network of similar depth, if trained directly using backpropagation, cannot acheive the tasks achieved using our layer wise training paradigm.

The C. elegans Connectome Consists of Homogenous Circuits with Defined Functional Roles

PubMed Central

Azulay, Aharon; Zaslaver, Alon

2016-01-01

A major goal of systems neuroscience is to decipher the structure-function relationship in neural networks. Here we study network functionality in light of the common-neighbor-rule (CNR) in which a pair of neurons is more likely to be connected the more common neighbors it shares. Focusing on the fully-mapped neural network of C. elegans worms, we establish that the CNR is an emerging property in this connectome. Moreover, sets of common neighbors form homogenous structures that appear in defined layers of the network. Simulations of signal propagation reveal their potential functional roles: signal amplification and short-term memory at the sensory/inter-neuron layer, and synchronized activity at the motoneuron layer supporting coordinated movement. A coarse-grained view of the neural network based on homogenous connected sets alone reveals a simple modular network architecture that is intuitive to understand. These findings provide a novel framework for analyzing larger, more complex, connectomes once these become available. PMID:27606684

Development and application of deep convolutional neural network in target detection

NASA Astrophysics Data System (ADS)

Jiang, Xiaowei; Wang, Chunping; Fu, Qiang

2018-04-01

With the development of big data and algorithms, deep convolution neural networks with more hidden layers have more powerful feature learning and feature expression ability than traditional machine learning methods, making artificial intelligence surpass human level in many fields. This paper first reviews the development and application of deep convolutional neural networks in the field of object detection in recent years, then briefly summarizes and ponders some existing problems in the current research, and the future development of deep convolutional neural network is prospected.

Genetic algorithm for neural networks optimization

NASA Astrophysics Data System (ADS)

Setyawati, Bina R.; Creese, Robert C.; Sahirman, Sidharta

2004-11-01

This paper examines the forecasting performance of multi-layer feed forward neural networks in modeling a particular foreign exchange rates, i.e. Japanese Yen/US Dollar. The effects of two learning methods, Back Propagation and Genetic Algorithm, in which the neural network topology and other parameters fixed, were investigated. The early results indicate that the application of this hybrid system seems to be well suited for the forecasting of foreign exchange rates. The Neural Networks and Genetic Algorithm were programmed using MATLAB«.

Spin switches for compact implementation of neuron and synapse

NASA Astrophysics Data System (ADS)

Quang Diep, Vinh; Sutton, Brian; Behin-Aein, Behtash; Datta, Supriyo

2014-06-01

Nanomagnets driven by spin currents provide a natural implementation for a neuron and a synapse: currents allow convenient summation of multiple inputs, while the magnet provides the threshold function. The objective of this paper is to explore the possibility of a hardware neural network implementation using a spin switch (SS) as its basic building block. SS is a recently proposed device based on established technology with a transistor-like gain and input-output isolation. This allows neural networks to be constructed with purely passive interconnections without intervening clocks or amplifiers. The weights for the neural network are conveniently adjusted through analog voltages that can be stored in a non-volatile manner in an underlying CMOS layer using a floating gate low dropout voltage regulator. The operation of a multi-layer SS neural network designed for character recognition is demonstrated using a standard simulation model based on coupled Landau-Lifshitz-Gilbert equations, one for each magnet in the network.

Study on algorithm of process neural network for soft sensing in sewage disposal system

NASA Astrophysics Data System (ADS)

Liu, Zaiwen; Xue, Hong; Wang, Xiaoyi; Yang, Bin; Lu, Siying

2006-11-01

A new method of soft sensing based on process neural network (PNN) for sewage disposal system is represented in the paper. PNN is an extension of traditional neural network, in which the inputs and outputs are time-variation. An aggregation operator is introduced to process neuron, and it makes the neuron network has the ability to deal with the information of space-time two dimensions at the same time, so the data processing enginery of biological neuron is imitated better than traditional neuron. Process neural network with the structure of three layers in which hidden layer is process neuron and input and output are common neurons for soft sensing is discussed. The intelligent soft sensing based on PNN may be used to fulfill measurement of the effluent BOD (Biochemical Oxygen Demand) from sewage disposal system, and a good training result of soft sensing was obtained by the method.

An improved advertising CTR prediction approach based on the fuzzy deep neural network

PubMed Central

Gao, Shu; Li, Mingjiang

2018-01-01

Combining a deep neural network with fuzzy theory, this paper proposes an advertising click-through rate (CTR) prediction approach based on a fuzzy deep neural network (FDNN). In this approach, fuzzy Gaussian-Bernoulli restricted Boltzmann machine (FGBRBM) is first applied to input raw data from advertising datasets. Next, fuzzy restricted Boltzmann machine (FRBM) is used to construct the fuzzy deep belief network (FDBN) with the unsupervised method layer by layer. Finally, fuzzy logistic regression (FLR) is utilized for modeling the CTR. The experimental results show that the proposed FDNN model outperforms several baseline models in terms of both data representation capability and robustness in advertising click log datasets with noise. PMID:29727443

An improved advertising CTR prediction approach based on the fuzzy deep neural network.

PubMed

Jiang, Zilong; Gao, Shu; Li, Mingjiang

2018-01-01

Combining a deep neural network with fuzzy theory, this paper proposes an advertising click-through rate (CTR) prediction approach based on a fuzzy deep neural network (FDNN). In this approach, fuzzy Gaussian-Bernoulli restricted Boltzmann machine (FGBRBM) is first applied to input raw data from advertising datasets. Next, fuzzy restricted Boltzmann machine (FRBM) is used to construct the fuzzy deep belief network (FDBN) with the unsupervised method layer by layer. Finally, fuzzy logistic regression (FLR) is utilized for modeling the CTR. The experimental results show that the proposed FDNN model outperforms several baseline models in terms of both data representation capability and robustness in advertising click log datasets with noise.

Artificial neural network based modelling approach for municipal solid waste gasification in a fluidized bed reactor.

PubMed

Pandey, Daya Shankar; Das, Saptarshi; Pan, Indranil; Leahy, James J; Kwapinski, Witold

2016-12-01

In this paper, multi-layer feed forward neural networks are used to predict the lower heating value of gas (LHV), lower heating value of gasification products including tars and entrained char (LHV p ) and syngas yield during gasification of municipal solid waste (MSW) during gasification in a fluidized bed reactor. These artificial neural networks (ANNs) with different architectures are trained using the Levenberg-Marquardt (LM) back-propagation algorithm and a cross validation is also performed to ensure that the results generalise to other unseen datasets. A rigorous study is carried out on optimally choosing the number of hidden layers, number of neurons in the hidden layer and activation function in a network using multiple Monte Carlo runs. Nine input and three output parameters are used to train and test various neural network architectures in both multiple output and single output prediction paradigms using the available experimental datasets. The model selection procedure is carried out to ascertain the best network architecture in terms of predictive accuracy. The simulation results show that the ANN based methodology is a viable alternative which can be used to predict the performance of a fluidized bed gasifier. Copyright © 2016 Elsevier Ltd. All rights reserved.

Convolutional Neural Network for Histopathological Analysis of Osteosarcoma.

PubMed

Mishra, Rashika; Daescu, Ovidiu; Leavey, Patrick; Rakheja, Dinesh; Sengupta, Anita

2018-03-01

Pathologists often deal with high complexity and sometimes disagreement over osteosarcoma tumor classification due to cellular heterogeneity in the dataset. Segmentation and classification of histology tissue in H&E stained tumor image datasets is a challenging task because of intra-class variations, inter-class similarity, crowded context, and noisy data. In recent years, deep learning approaches have led to encouraging results in breast cancer and prostate cancer analysis. In this article, we propose convolutional neural network (CNN) as a tool to improve efficiency and accuracy of osteosarcoma tumor classification into tumor classes (viable tumor, necrosis) versus nontumor. The proposed CNN architecture contains eight learned layers: three sets of stacked two convolutional layers interspersed with max pooling layers for feature extraction and two fully connected layers with data augmentation strategies to boost performance. The use of a neural network results in higher accuracy of average 92% for the classification. We compare the proposed architecture with three existing and proven CNN architectures for image classification: AlexNet, LeNet, and VGGNet. We also provide a pipeline to calculate percentage necrosis in a given whole slide image. We conclude that the use of neural networks can assure both high accuracy and efficiency in osteosarcoma classification.

Traffic sign recognition based on deep convolutional neural network

NASA Astrophysics Data System (ADS)

Yin, Shi-hao; Deng, Ji-cai; Zhang, Da-wei; Du, Jing-yuan

2017-11-01

Traffic sign recognition (TSR) is an important component of automated driving systems. It is a rather challenging task to design a high-performance classifier for the TSR system. In this paper, we propose a new method for TSR system based on deep convolutional neural network. In order to enhance the expression of the network, a novel structure (dubbed block-layer below) which combines network-in-network and residual connection is designed. Our network has 10 layers with parameters (block-layer seen as a single layer): the first seven are alternate convolutional layers and block-layers, and the remaining three are fully-connected layers. We train our TSR network on the German traffic sign recognition benchmark (GTSRB) dataset. To reduce overfitting, we perform data augmentation on the training images and employ a regularization method named "dropout". The activation function we employ in our network adopts scaled exponential linear units (SELUs), which can induce self-normalizing properties. To speed up the training, we use an efficient GPU to accelerate the convolutional operation. On the test dataset of GTSRB, we achieve the accuracy rate of 99.67%, exceeding the state-of-the-art results.

Reducing neural network training time with parallel processing

NASA Technical Reports Server (NTRS)

Rogers, James L., Jr.; Lamarsh, William J., II

1995-01-01

Obtaining optimal solutions for engineering design problems is often expensive because the process typically requires numerous iterations involving analysis and optimization programs. Previous research has shown that a near optimum solution can be obtained in less time by simulating a slow, expensive analysis with a fast, inexpensive neural network. A new approach has been developed to further reduce this time. This approach decomposes a large neural network into many smaller neural networks that can be trained in parallel. Guidelines are developed to avoid some of the pitfalls when training smaller neural networks in parallel. These guidelines allow the engineer: to determine the number of nodes on the hidden layer of the smaller neural networks; to choose the initial training weights; and to select a network configuration that will capture the interactions among the smaller neural networks. This paper presents results describing how these guidelines are developed.

[Early warning on measles through the neural networks].

PubMed

Yu, Bin; Ding, Chun; Wei, Shan-bo; Chen, Bang-hua; Liu, Pu-lin; Luo, Tong-yong; Wang, Jia-gang; Pan, Zhi-wei; Lu, Jun-an

2011-01-01

To discuss the effects on early warning of measles, using the neural networks. Based on the available data through monthly and weekly reports on measles from January 1986 to August 2006 in Wuhan city. The modal was developed using the neural networks to predict and analyze the prevalence and incidence of measles. When the dynamic time series modal was established with back propagation (BP) networks consisting of two layers, if p was assigned as 9, the convergence speed was acceptable and the correlation coefficient was equal to 0.85. It was more acceptable for monthly forecasting the specific value, but better for weekly forecasting the classification under probabilistic neural networks (PNN). When data was big enough to serve the purpose, it seemed more feasible for early warning using the two-layer BP networks. However, when data was not enough, then PNN could be used for the purpose of prediction. This method seemed feasible to be used in the system for early warning.

Temporal neural networks and transient analysis of complex engineering systems

NASA Astrophysics Data System (ADS)

Uluyol, Onder

A theory is introduced for a multi-layered Local Output Gamma Feedback (LOGF) neural network within the paradigm of Locally-Recurrent Globally-Feedforward neural networks. It is developed for the identification, prediction, and control tasks of spatio-temporal systems and allows for the presentation of different time scales through incorporation of a gamma memory. It is initially applied to the tasks of sunspot and Mackey-Glass series prediction as benchmarks, then it is extended to the task of power level control of a nuclear reactor at different fuel cycle conditions. The developed LOGF neuron model can also be viewed as a Transformed Input and State (TIS) Gamma memory for neural network architectures for temporal processing. The novel LOGF neuron model extends the static neuron model by incorporating into it a short-term memory structure in the form of a digital gamma filter. A feedforward neural network made up of LOGF neurons can thus be used to model dynamic systems. A learning algorithm based upon the Backpropagation-Through-Time (BTT) approach is derived. It is applicable for training a general L-layer LOGF neural network. The spatial and temporal weights and parameters of the network are iteratively optimized for a given problem using the derived learning algorithm.

Self-growing neural network architecture using crisp and fuzzy entropy

NASA Technical Reports Server (NTRS)

Cios, Krzysztof J.

1992-01-01

The paper briefly describes the self-growing neural network algorithm, CID2, which makes decision trees equivalent to hidden layers of a neural network. The algorithm generates a feedforward architecture using crisp and fuzzy entropy measures. The results of a real-life recognition problem of distinguishing defects in a glass ribbon and of a benchmark problem of differentiating two spirals are shown and discussed.

Self-growing neural network architecture using crisp and fuzzy entropy

NASA Technical Reports Server (NTRS)

Cios, Krzysztof J.

1992-01-01

The paper briefly describes the self-growing neural network algorithm, CID3, which makes decision trees equivalent to hidden layers of a neural network. The algorithm generates a feedforward architecture using crisp and fuzzy entropy measures. The results for a real-life recognition problem of distinguishing defects in a glass ribbon, and for a benchmark problen of telling two spirals apart are shown and discussed.

Neural-Network-Development Program

NASA Technical Reports Server (NTRS)

Phillips, Todd A.

1993-01-01

NETS, software tool for development and evaluation of neural networks, provides simulation of neural-network algorithms plus computing environment for development of such algorithms. Uses back-propagation learning method for all of networks it creates. Enables user to customize patterns of connections between layers of network. Also provides features for saving, during learning process, values of weights, providing more-precise control over learning process. Written in ANSI standard C language. Machine-independent version (MSC-21588) includes only code for command-line-interface version of NETS 3.0.

Predicting cloud-to-ground lightning with neural networks

NASA Technical Reports Server (NTRS)

Barnes, Arnold A., Jr.; Frankel, Donald; Draper, James Stark

1991-01-01

A neural network is being trained to predict lightning at Cape Canaveral for periods up to two hours in advance. Inputs consist of ground based field mill data, meteorological tower data, lightning location data, and radiosonde data. High values of the field mill data and rapid changes in the field mill data, offset in time, provide the forecasts or desired output values used to train the neural network through backpropagation. Examples of input data are shown and an example of data compression using a hidden layer in the neural network is discussed.

Design of a Thermoacoustic Sensor for Low Intensity Ultrasound Measurements Based on an Artificial Neural Network.

PubMed

Xing, Jida; Chen, Jie

2015-06-23

In therapeutic ultrasound applications, accurate ultrasound output intensities are crucial because the physiological effects of therapeutic ultrasound are very sensitive to the intensity and duration of these applications. Although radiation force balance is a benchmark technique for measuring ultrasound intensity and power, it is costly, difficult to operate, and compromised by noise vibration. To overcome these limitations, the development of a low-cost, easy to operate, and vibration-resistant alternative device is necessary for rapid ultrasound intensity measurement. Therefore, we proposed and validated a novel two-layer thermoacoustic sensor using an artificial neural network technique to accurately measure low ultrasound intensities between 30 and 120 mW/cm2. The first layer of the sensor design is a cylindrical absorber made of plexiglass, followed by a second layer composed of polyurethane rubber with a high attenuation coefficient to absorb extra ultrasound energy. The sensor determined ultrasound intensities according to a temperature elevation induced by heat converted from incident acoustic energy. Compared with our previous one-layer sensor design, the new two-layer sensor enhanced the ultrasound absorption efficiency to provide more rapid and reliable measurements. Using a three-dimensional model in the K-wave toolbox, our simulation of the ultrasound propagation process demonstrated that the two-layer design is more efficient than the single layer design. We also integrated an artificial neural network algorithm to compensate for the large measurement offset. After obtaining multiple parameters of the sensor characteristics through calibration, the artificial neural network is built to correct temperature drifts and increase the reliability of our thermoacoustic measurements through iterative training about ten seconds. The performance of the artificial neural network method was validated through a series of experiments. Compared to our previous design, the new design reduced sensing time from 20 s to 12 s, and the sensor's average error from 3.97 mW/cm2 to 1.31 mW/cm2 respectively.

Design of a Thermoacoustic Sensor for Low Intensity Ultrasound Measurements Based on an Artificial Neural Network

PubMed Central

Xing, Jida; Chen, Jie

2015-01-01

In therapeutic ultrasound applications, accurate ultrasound output intensities are crucial because the physiological effects of therapeutic ultrasound are very sensitive to the intensity and duration of these applications. Although radiation force balance is a benchmark technique for measuring ultrasound intensity and power, it is costly, difficult to operate, and compromised by noise vibration. To overcome these limitations, the development of a low-cost, easy to operate, and vibration-resistant alternative device is necessary for rapid ultrasound intensity measurement. Therefore, we proposed and validated a novel two-layer thermoacoustic sensor using an artificial neural network technique to accurately measure low ultrasound intensities between 30 and 120 mW/cm2. The first layer of the sensor design is a cylindrical absorber made of plexiglass, followed by a second layer composed of polyurethane rubber with a high attenuation coefficient to absorb extra ultrasound energy. The sensor determined ultrasound intensities according to a temperature elevation induced by heat converted from incident acoustic energy. Compared with our previous one-layer sensor design, the new two-layer sensor enhanced the ultrasound absorption efficiency to provide more rapid and reliable measurements. Using a three-dimensional model in the K-wave toolbox, our simulation of the ultrasound propagation process demonstrated that the two-layer design is more efficient than the single layer design. We also integrated an artificial neural network algorithm to compensate for the large measurement offset. After obtaining multiple parameters of the sensor characteristics through calibration, the artificial neural network is built to correct temperature drifts and increase the reliability of our thermoacoustic measurements through iterative training about ten seconds. The performance of the artificial neural network method was validated through a series of experiments. Compared to our previous design, the new design reduced sensing time from 20 s to 12 s, and the sensor’s average error from 3.97 mW/cm2 to 1.31 mW/cm2 respectively. PMID:26110412

Numerical Analysis of Modeling Based on Improved Elman Neural Network

PubMed Central

Jie, Shao

2014-01-01

A modeling based on the improved Elman neural network (IENN) is proposed to analyze the nonlinear circuits with the memory effect. The hidden layer neurons are activated by a group of Chebyshev orthogonal basis functions instead of sigmoid functions in this model. The error curves of the sum of squared error (SSE) varying with the number of hidden neurons and the iteration step are studied to determine the number of the hidden layer neurons. Simulation results of the half-bridge class-D power amplifier (CDPA) with two-tone signal and broadband signals as input have shown that the proposed behavioral modeling can reconstruct the system of CDPAs accurately and depict the memory effect of CDPAs well. Compared with Volterra-Laguerre (VL) model, Chebyshev neural network (CNN) model, and basic Elman neural network (BENN) model, the proposed model has better performance. PMID:25054172

Unsupervised discrimination of patterns in spiking neural networks with excitatory and inhibitory synaptic plasticity

PubMed Central

Srinivasa, Narayan; Cho, Youngkwan

2014-01-01

A spiking neural network model is described for learning to discriminate among spatial patterns in an unsupervised manner. The network anatomy consists of source neurons that are activated by external inputs, a reservoir that resembles a generic cortical layer with an excitatory-inhibitory (EI) network and a sink layer of neurons for readout. Synaptic plasticity in the form of STDP is imposed on all the excitatory and inhibitory synapses at all times. While long-term excitatory STDP enables sparse and efficient learning of the salient features in inputs, inhibitory STDP enables this learning to be stable by establishing a balance between excitatory and inhibitory currents at each neuron in the network. The synaptic weights between source and reservoir neurons form a basis set for the input patterns. The neural trajectories generated in the reservoir due to input stimulation and lateral connections between reservoir neurons can be readout by the sink layer neurons. This activity is used for adaptation of synapses between reservoir and sink layer neurons. A new measure called the discriminability index (DI) is introduced to compute if the network can discriminate between old patterns already presented in an initial training session. The DI is also used to compute if the network adapts to new patterns without losing its ability to discriminate among old patterns. The final outcome is that the network is able to correctly discriminate between all patterns—both old and new. This result holds as long as inhibitory synapses employ STDP to continuously enable current balance in the network. The results suggest a possible direction for future investigation into how spiking neural networks could address the stability-plasticity question despite having continuous synaptic plasticity. PMID:25566045

Unsupervised discrimination of patterns in spiking neural networks with excitatory and inhibitory synaptic plasticity.

PubMed

Srinivasa, Narayan; Cho, Youngkwan

2014-01-01

A spiking neural network model is described for learning to discriminate among spatial patterns in an unsupervised manner. The network anatomy consists of source neurons that are activated by external inputs, a reservoir that resembles a generic cortical layer with an excitatory-inhibitory (EI) network and a sink layer of neurons for readout. Synaptic plasticity in the form of STDP is imposed on all the excitatory and inhibitory synapses at all times. While long-term excitatory STDP enables sparse and efficient learning of the salient features in inputs, inhibitory STDP enables this learning to be stable by establishing a balance between excitatory and inhibitory currents at each neuron in the network. The synaptic weights between source and reservoir neurons form a basis set for the input patterns. The neural trajectories generated in the reservoir due to input stimulation and lateral connections between reservoir neurons can be readout by the sink layer neurons. This activity is used for adaptation of synapses between reservoir and sink layer neurons. A new measure called the discriminability index (DI) is introduced to compute if the network can discriminate between old patterns already presented in an initial training session. The DI is also used to compute if the network adapts to new patterns without losing its ability to discriminate among old patterns. The final outcome is that the network is able to correctly discriminate between all patterns-both old and new. This result holds as long as inhibitory synapses employ STDP to continuously enable current balance in the network. The results suggest a possible direction for future investigation into how spiking neural networks could address the stability-plasticity question despite having continuous synaptic plasticity.

Feature to prototype transition in neural networks

NASA Astrophysics Data System (ADS)

Krotov, Dmitry; Hopfield, John

Models of associative memory with higher order (higher than quadratic) interactions, and their relationship to neural networks used in deep learning are discussed. Associative memory is conventionally described by recurrent neural networks with dynamical convergence to stable points. Deep learning typically uses feedforward neural nets without dynamics. However, a simple duality relates these two different views when applied to problems of pattern classification. From the perspective of associative memory such models deserve attention because they make it possible to store a much larger number of memories, compared to the quadratic case. In the dual description, these models correspond to feedforward neural networks with one hidden layer and unusual activation functions transmitting the activities of the visible neurons to the hidden layer. These activation functions are rectified polynomials of a higher degree rather than the rectified linear functions used in deep learning. The network learns representations of the data in terms of features for rectified linear functions, but as the power in the activation function is increased there is a gradual shift to a prototype-based representation, the two extreme regimes of pattern recognition known in cognitive psychology. Simons Center for Systems Biology.

Adaptive neural network motion control of manipulators with experimental evaluations.

PubMed

Puga-Guzmán, S; Moreno-Valenzuela, J; Santibáñez, V

2014-01-01

A nonlinear proportional-derivative controller plus adaptive neuronal network compensation is proposed. With the aim of estimating the desired torque, a two-layer neural network is used. Then, adaptation laws for the neural network weights are derived. Asymptotic convergence of the position and velocity tracking errors is proven, while the neural network weights are shown to be uniformly bounded. The proposed scheme has been experimentally validated in real time. These experimental evaluations were carried in two different mechanical systems: a horizontal two degrees-of-freedom robot and a vertical one degree-of-freedom arm which is affected by the gravitational force. In each one of the two experimental set-ups, the proposed scheme was implemented without and with adaptive neural network compensation. Experimental results confirmed the tracking accuracy of the proposed adaptive neural network-based controller.

Adaptive Neural Network Motion Control of Manipulators with Experimental Evaluations

PubMed Central

Puga-Guzmán, S.; Moreno-Valenzuela, J.; Santibáñez, V.

2014-01-01

A nonlinear proportional-derivative controller plus adaptive neuronal network compensation is proposed. With the aim of estimating the desired torque, a two-layer neural network is used. Then, adaptation laws for the neural network weights are derived. Asymptotic convergence of the position and velocity tracking errors is proven, while the neural network weights are shown to be uniformly bounded. The proposed scheme has been experimentally validated in real time. These experimental evaluations were carried in two different mechanical systems: a horizontal two degrees-of-freedom robot and a vertical one degree-of-freedom arm which is affected by the gravitational force. In each one of the two experimental set-ups, the proposed scheme was implemented without and with adaptive neural network compensation. Experimental results confirmed the tracking accuracy of the proposed adaptive neural network-based controller. PMID:24574910

An artificial neural network model for periodic trajectory generation

NASA Astrophysics Data System (ADS)

Shankar, S.; Gander, R. E.; Wood, H. C.

A neural network model based on biological systems was developed for potential robotic application. The model consists of three interconnected layers of artificial neurons or units: an input layer subdivided into state and plan units, an output layer, and a hidden layer between the two outer layers which serves to implement nonlinear mappings between the input and output activation vectors. Weighted connections are created between the three layers, and learning is effected by modifying these weights. Feedback connections between the output and the input state serve to make the network operate as a finite state machine. The activation vector of the plan units of the input layer emulates the supraspinal commands in biological central pattern generators in that different plan activation vectors correspond to different sequences or trajectories being recalled, even with different frequencies. Three trajectories were chosen for implementation, and learning was accomplished in 10,000 trials. The fault tolerant behavior, adaptiveness, and phase maintenance of the implemented network are discussed.

Spatial-Temporal Reasoning Applications of Computational Intelligence in the Game of Go and Computer Networks

DTIC Science & Technology

2012-01-01

dimensionality, Tesauro used a backpropagation- based , three-layer neural network and implemented the outcome from a self-play game as the reinforcement signal...a school of fish, flock of birds, and colony of ants. Our literature review reveals that no one has used PSO to train the neural network ...trained with a variant of PSO called cellular PSO (CPSO). CSRN is a supervised learning neural network (SLNN). The proposed algorithm for the

Adiabatic superconducting cells for ultra-low-power artificial neural networks.

PubMed

Schegolev, Andrey E; Klenov, Nikolay V; Soloviev, Igor I; Tereshonok, Maxim V

2016-01-01

We propose the concept of using superconducting quantum interferometers for the implementation of neural network algorithms with extremely low power dissipation. These adiabatic elements are Josephson cells with sigmoid- and Gaussian-like activation functions. We optimize their parameters for application in three-layer perceptron and radial basis function networks.

A biologically inspired neural network for dynamic programming.

PubMed

Francelin Romero, R A; Kacpryzk, J; Gomide, F

2001-12-01

An artificial neural network with a two-layer feedback topology and generalized recurrent neurons, for solving nonlinear discrete dynamic optimization problems, is developed. A direct method to assign the weights of neural networks is presented. The method is based on Bellmann's Optimality Principle and on the interchange of information which occurs during the synaptic chemical processing among neurons. The neural network based algorithm is an advantageous approach for dynamic programming due to the inherent parallelism of the neural networks; further it reduces the severity of computational problems that can occur in methods like conventional methods. Some illustrative application examples are presented to show how this approach works out including the shortest path and fuzzy decision making problems.

Neural computation of arithmetic functions

NASA Technical Reports Server (NTRS)

Siu, Kai-Yeung; Bruck, Jehoshua

1990-01-01

An area of application of neural networks is considered. A neuron is modeled as a linear threshold gate, and the network architecture considered is the layered feedforward network. It is shown how common arithmetic functions such as multiplication and sorting can be efficiently computed in a shallow neural network. Some known results are improved by showing that the product of two n-bit numbers and sorting of n n-bit numbers can be computed by a polynomial-size neural network using only four and five unit delays, respectively. Moreover, the weights of each threshold element in the neural networks require O(log n)-bit (instead of n-bit) accuracy. These results can be extended to more complicated functions such as multiple products, division, rational functions, and approximation of analytic functions.

Using deep recurrent neural network for direct beam solar irradiance cloud screening

NASA Astrophysics Data System (ADS)

Chen, Maosi; Davis, John M.; Liu, Chaoshun; Sun, Zhibin; Zempila, Melina Maria; Gao, Wei

2017-09-01

Cloud screening is an essential procedure for in-situ calibration and atmospheric properties retrieval on (UV-)MultiFilter Rotating Shadowband Radiometer [(UV-)MFRSR]. Previous study has explored a cloud screening algorithm for direct-beam (UV-)MFRSR voltage measurements based on the stability assumption on a long time period (typically a half day or a whole day). To design such an algorithm requires in-depth understanding of radiative transfer and delicate data manipulation. Recent rapid developments on deep neural network and computation hardware have opened a window for modeling complicated End-to-End systems with a standardized strategy. In this study, a multi-layer dynamic bidirectional recurrent neural network is built for determining the cloudiness on each time point with a 17-year training dataset and tested with another 1-year dataset. The dataset is the daily 3-minute cosine corrected voltages, airmasses, and the corresponding cloud/clear-sky labels at two stations of the USDA UV-B Monitoring and Research Program. The results show that the optimized neural network model (3-layer, 250 hidden units, and 80 epochs of training) has an overall test accuracy of 97.87% (97.56% for the Oklahoma site and 98.16% for the Hawaii site). Generally, the neural network model grasps the key concept of the original model to use data in the entire day rather than short nearby measurements to perform cloud screening. A scrutiny of the logits layer suggests that the neural network model automatically learns a way to calculate a quantity similar to total optical depth and finds an appropriate threshold for cloud screening.

Using Hybrid Algorithm to Improve Intrusion Detection in Multi Layer Feed Forward Neural Networks

ERIC Educational Resources Information Center

Ray, Loye Lynn

2014-01-01

The need for detecting malicious behavior on a computer networks continued to be important to maintaining a safe and secure environment. The purpose of this study was to determine the relationship of multilayer feed forward neural network architecture to the ability of detecting abnormal behavior in networks. This involved building, training, and…

Spin switches for compact implementation of neuron and synapse

DOE Office of Scientific and Technical Information (OSTI.GOV)

Quang Diep, Vinh, E-mail: vdiep@purdue.edu; Sutton, Brian; Datta, Supriyo

2014-06-02

Nanomagnets driven by spin currents provide a natural implementation for a neuron and a synapse: currents allow convenient summation of multiple inputs, while the magnet provides the threshold function. The objective of this paper is to explore the possibility of a hardware neural network implementation using a spin switch (SS) as its basic building block. SS is a recently proposed device based on established technology with a transistor-like gain and input-output isolation. This allows neural networks to be constructed with purely passive interconnections without intervening clocks or amplifiers. The weights for the neural network are conveniently adjusted through analog voltagesmore » that can be stored in a non-volatile manner in an underlying CMOS layer using a floating gate low dropout voltage regulator. The operation of a multi-layer SS neural network designed for character recognition is demonstrated using a standard simulation model based on coupled Landau-Lifshitz-Gilbert equations, one for each magnet in the network.« less

The use of global image characteristics for neural network pattern recognitions

NASA Astrophysics Data System (ADS)

Kulyas, Maksim O.; Kulyas, Oleg L.; Loshkarev, Aleksey S.

2017-04-01

The recognition system is observed, where the information is transferred by images of symbols generated by a television camera. For descriptors of objects the coefficients of two-dimensional Fourier transformation generated in a special way. For solution of the task of classification the one-layer neural network trained on reference images is used. Fast learning of a neural network with a single neuron calculation of coefficients is applied.

Automated road marking recognition system

NASA Astrophysics Data System (ADS)

Ziyatdinov, R. R.; Shigabiev, R. R.; Talipov, D. N.

2017-09-01

Development of the automated road marking recognition systems in existing and future vehicles control systems is an urgent task. One way to implement such systems is the use of neural networks. To test the possibility of using neural network software has been developed with the use of a single-layer perceptron. The resulting system based on neural network has successfully coped with the task both when driving in the daytime and at night.

A Combined Adaptive Neural Network and Nonlinear Model Predictive Control for Multirate Networked Industrial Process Control.

PubMed

Wang, Tong; Gao, Huijun; Qiu, Jianbin

2016-02-01

This paper investigates the multirate networked industrial process control problem in double-layer architecture. First, the output tracking problem for sampled-data nonlinear plant at device layer with sampling period T(d) is investigated using adaptive neural network (NN) control, and it is shown that the outputs of subsystems at device layer can track the decomposed setpoints. Then, the outputs and inputs of the device layer subsystems are sampled with sampling period T(u) at operation layer to form the index prediction, which is used to predict the overall performance index at lower frequency. Radial basis function NN is utilized as the prediction function due to its approximation ability. Then, considering the dynamics of the overall closed-loop system, nonlinear model predictive control method is proposed to guarantee the system stability and compensate the network-induced delays and packet dropouts. Finally, a continuous stirred tank reactor system is given in the simulation part to demonstrate the effectiveness of the proposed method.

Neural Network Prediction of Aluminum-Lithium Weld Strengths from Acoustic Emission Amplitude Data

NASA Technical Reports Server (NTRS)

Hill, Eric v. K.; Israel, Peggy L.; Knotts, Gregory L.

1993-01-01

Acoustic Emission (AE) flaw growth activity was monitored in aluminum-lithium weld specimens from the onset tensile loading to failure. Data on actual ultimate strengths together with AE data from the beginning of loading up to 25 percent of the expected ultimate strength were used to train a backpropagation neural network to predict ultimate strengths. Architecturally, the fully interconnected network consisted of an input layer for the AE amplitude data, a hidden layer to accommodate failure mechanism mapping, and an output layer for ultimate strength prediction. The trained network was the applied to the prediction of ultimate strengths in the remaining six specimens. The worst case prediction error was found to be +2.6 percent.

Study on pattern recognition of Raman spectrum based on fuzzy neural network

NASA Astrophysics Data System (ADS)

Zheng, Xiangxiang; Lv, Xiaoyi; Mo, Jiaqing

2017-10-01

Hydatid disease is a serious parasitic disease in many regions worldwide, especially in Xinjiang, China. Raman spectrum of the serum of patients with echinococcosis was selected as the research object in this paper. The Raman spectrum of blood samples from healthy people and patients with echinococcosis are measured, of which the spectrum characteristics are analyzed. The fuzzy neural network not only has the ability of fuzzy logic to deal with uncertain information, but also has the ability to store knowledge of neural network, so it is combined with the Raman spectrum on the disease diagnosis problem based on Raman spectrum. Firstly, principal component analysis (PCA) is used to extract the principal components of the Raman spectrum, reducing the network input and accelerating the prediction speed and accuracy of Network based on remaining the original data. Then, the information of the extracted principal component is used as the input of the neural network, the hidden layer of the network is the generation of rules and the inference process, and the output layer of the network is fuzzy classification output. Finally, a part of samples are randomly selected for the use of training network, then the trained network is used for predicting the rest of the samples, and the predicted results are compared with general BP neural network to illustrate the feasibility and advantages of fuzzy neural network. Success in this endeavor would be helpful for the research work of spectroscopic diagnosis of disease and it can be applied in practice in many other spectral analysis technique fields.

A Data-Driven Response Virtual Sensor Technique with Partial Vibration Measurements Using Convolutional Neural Network.

PubMed

Sun, Shan-Bin; He, Yuan-Yuan; Zhou, Si-Da; Yue, Zhen-Jiang

2017-12-12

Measurement of dynamic responses plays an important role in structural health monitoring, damage detection and other fields of research. However, in aerospace engineering, the physical sensors are limited in the operational conditions of spacecraft, due to the severe environment in outer space. This paper proposes a virtual sensor model with partial vibration measurements using a convolutional neural network. The transmissibility function is employed as prior knowledge. A four-layer neural network with two convolutional layers, one fully connected layer, and an output layer is proposed as the predicting model. Numerical examples of two different structural dynamic systems demonstrate the performance of the proposed approach. The excellence of the novel technique is further indicated using a simply supported beam experiment comparing to a modal-model-based virtual sensor, which uses modal parameters, such as mode shapes, for estimating the responses of the faulty sensors. The results show that the presented data-driven response virtual sensor technique can predict structural response with high accuracy.

A Data-Driven Response Virtual Sensor Technique with Partial Vibration Measurements Using Convolutional Neural Network

PubMed Central

Sun, Shan-Bin; He, Yuan-Yuan; Zhou, Si-Da; Yue, Zhen-Jiang

2017-01-01

Measurement of dynamic responses plays an important role in structural health monitoring, damage detection and other fields of research. However, in aerospace engineering, the physical sensors are limited in the operational conditions of spacecraft, due to the severe environment in outer space. This paper proposes a virtual sensor model with partial vibration measurements using a convolutional neural network. The transmissibility function is employed as prior knowledge. A four-layer neural network with two convolutional layers, one fully connected layer, and an output layer is proposed as the predicting model. Numerical examples of two different structural dynamic systems demonstrate the performance of the proposed approach. The excellence of the novel technique is further indicated using a simply supported beam experiment comparing to a modal-model-based virtual sensor, which uses modal parameters, such as mode shapes, for estimating the responses of the faulty sensors. The results show that the presented data-driven response virtual sensor technique can predict structural response with high accuracy. PMID:29231868

Neural networks for vertical microcode compaction

NASA Astrophysics Data System (ADS)

Chu, Pong P.

1992-09-01

Neural networks provide an alternative way to solve complex optimization problems. Instead of performing a program of instructions sequentially as in a traditional computer, neural network model explores many competing hypotheses simultaneously using its massively parallel net. The paper shows how to use the neural network approach to perform vertical micro-code compaction for a micro-programmed control unit. The compaction procedure includes two basic steps. The first step determines the compatibility classes and the second step selects a minimal subset to cover the control signals. Since the selection process is an NP- complete problem, to find an optimal solution is impractical. In this study, we employ a customized neural network to obtain the minimal subset. We first formalize this problem, and then define an `energy function' and map it to a two-layer fully connected neural network. The modified network has two types of neurons and can always obtain a valid solution.

Predicting Item Difficulty in a Reading Comprehension Test with an Artificial Neural Network.

ERIC Educational Resources Information Center

Perkins, Kyle; And Others

1995-01-01

This article reports the results of using a three-layer back propagation artificial neural network to predict item difficulty in a reading comprehension test. Three classes of variables were examined: text structure, propositional analysis, and cognitive demand. Results demonstrate that the networks can consistently predict item difficulty. (JL)

Character Recognition Using Genetically Trained Neural Networks

DOE Office of Scientific and Technical Information (OSTI.GOV)

Diniz, C.; Stantz, K.M.; Trahan, M.W.

1998-10-01

Computationally intelligent recognition of characters and symbols addresses a wide range of applications including foreign language translation and chemical formula identification. The combination of intelligent learning and optimization algorithms with layered neural structures offers powerful techniques for character recognition. These techniques were originally developed by Sandia National Laboratories for pattern and spectral analysis; however, their ability to optimize vast amounts of data make them ideal for character recognition. An adaptation of the Neural Network Designer soflsvare allows the user to create a neural network (NN_) trained by a genetic algorithm (GA) that correctly identifies multiple distinct characters. The initial successfidmore » recognition of standard capital letters can be expanded to include chemical and mathematical symbols and alphabets of foreign languages, especially Arabic and Chinese. The FIN model constructed for this project uses a three layer feed-forward architecture. To facilitate the input of characters and symbols, a graphic user interface (GUI) has been developed to convert the traditional representation of each character or symbol to a bitmap. The 8 x 8 bitmap representations used for these tests are mapped onto the input nodes of the feed-forward neural network (FFNN) in a one-to-one correspondence. The input nodes feed forward into a hidden layer, and the hidden layer feeds into five output nodes correlated to possible character outcomes. During the training period the GA optimizes the weights of the NN until it can successfully recognize distinct characters. Systematic deviations from the base design test the network's range of applicability. Increasing capacity, the number of letters to be recognized, requires a nonlinear increase in the number of hidden layer neurodes. Optimal character recognition performance necessitates a minimum threshold for the number of cases when genetically training the net. And, the amount of noise significantly degrades character recognition efficiency, some of which can be overcome by adding noise during training and optimizing the form of the network's activation fimction.« less

Financial time series prediction using spiking neural networks.

PubMed

Reid, David; Hussain, Abir Jaafar; Tawfik, Hissam

2014-01-01

In this paper a novel application of a particular type of spiking neural network, a Polychronous Spiking Network, was used for financial time series prediction. It is argued that the inherent temporal capabilities of this type of network are suited to non-stationary data such as this. The performance of the spiking neural network was benchmarked against three systems: two "traditional", rate-encoded, neural networks; a Multi-Layer Perceptron neural network and a Dynamic Ridge Polynomial neural network, and a standard Linear Predictor Coefficients model. For this comparison three non-stationary and noisy time series were used: IBM stock data; US/Euro exchange rate data, and the price of Brent crude oil. The experiments demonstrated favourable prediction results for the Spiking Neural Network in terms of Annualised Return and prediction error for 5-Step ahead predictions. These results were also supported by other relevant metrics such as Maximum Drawdown and Signal-To-Noise ratio. This work demonstrated the applicability of the Polychronous Spiking Network to financial data forecasting and this in turn indicates the potential of using such networks over traditional systems in difficult to manage non-stationary environments.

LavaNet—Neural network development environment in a general mine planning package

NASA Astrophysics Data System (ADS)

Kapageridis, Ioannis Konstantinou; Triantafyllou, A. G.

2011-04-01

LavaNet is a series of scripts written in Perl that gives access to a neural network simulation environment inside a general mine planning package. A well known and a very popular neural network development environment, the Stuttgart Neural Network Simulator, is used as the base for the development of neural networks. LavaNet runs inside VULCAN™—a complete mine planning package with advanced database, modelling and visualisation capabilities. LavaNet is taking advantage of VULCAN's Perl based scripting environment, Lava, to bring all the benefits of neural network development and application to geologists, mining engineers and other users of the specific mine planning package. LavaNet enables easy development of neural network training data sets using information from any of the data and model structures available, such as block models and drillhole databases. Neural networks can be trained inside VULCAN™ and the results be used to generate new models that can be visualised in 3D. Direct comparison of developed neural network models with conventional and geostatistical techniques is now possible within the same mine planning software package. LavaNet supports Radial Basis Function networks, Multi-Layer Perceptrons and Self-Organised Maps.

Development of Artificial Neural Network Model for Diesel Fuel Properties Prediction using Vibrational Spectroscopy.

PubMed

Bolanča, Tomislav; Marinović, Slavica; Ukić, Sime; Jukić, Ante; Rukavina, Vinko

2012-06-01

This paper describes development of artificial neural network models which can be used to correlate and predict diesel fuel properties from several FTIR-ATR absorbances and Raman intensities as input variables. Multilayer feed forward and radial basis function neural networks have been used to rapid and simultaneous prediction of cetane number, cetane index, density, viscosity, distillation temperatures at 10% (T10), 50% (T50) and 90% (T90) recovery, contents of total aromatics and polycyclic aromatic hydrocarbons of commercial diesel fuels. In this study two-phase training procedures for multilayer feed forward networks were applied. While first phase training algorithm was constantly the back propagation one, two second phase training algorithms were varied and compared, namely: conjugate gradient and quasi Newton. In case of radial basis function network, radial layer was trained using K-means radial assignment algorithm and three different radial spread algorithms: explicit, isotropic and K-nearest neighbour. The number of hidden layer neurons and experimental data points used for the training set have been optimized for both neural networks in order to insure good predictive ability by reducing unnecessary experimental work. This work shows that developed artificial neural network models can determine main properties of diesel fuels simultaneously based on a single and fast IR or Raman measurement.

A review and analysis of neural networks for classification of remotely sensed multispectral imagery

NASA Technical Reports Server (NTRS)

Paola, Justin D.; Schowengerdt, Robert A.

1993-01-01

A literature survey and analysis of the use of neural networks for the classification of remotely sensed multispectral imagery is presented. As part of a brief mathematical review, the backpropagation algorithm, which is the most common method of training multi-layer networks, is discussed with an emphasis on its application to pattern recognition. The analysis is divided into five aspects of neural network classification: (1) input data preprocessing, structure, and encoding; (2) output encoding and extraction of classes; (3) network architecture, (4) training algorithms; and (5) comparisons to conventional classifiers. The advantages of the neural network method over traditional classifiers are its non-parametric nature, arbitrary decision boundary capabilities, easy adaptation to different types of data and input structures, fuzzy output values that can enhance classification, and good generalization for use with multiple images. The disadvantages of the method are slow training time, inconsistent results due to random initial weights, and the requirement of obscure initialization values (e.g., learning rate and hidden layer size). Possible techniques for ameliorating these problems are discussed. It is concluded that, although the neural network method has several unique capabilities, it will become a useful tool in remote sensing only if it is made faster, more predictable, and easier to use.

A neural network model for credit risk evaluation.

PubMed

Khashman, Adnan

2009-08-01

Credit scoring is one of the key analytical techniques in credit risk evaluation which has been an active research area in financial risk management. This paper presents a credit risk evaluation system that uses a neural network model based on the back propagation learning algorithm. We train and implement the neural network to decide whether to approve or reject a credit application, using seven learning schemes and real world credit applications from the Australian credit approval datasets. A comparison of the system performance under the different learning schemes is provided, furthermore, we compare the performance of two neural networks; with one and two hidden layers following the ideal learning scheme. Experimental results suggest that neural networks can be effectively used in automatic processing of credit applications.

Character recognition from trajectory by recurrent spiking neural networks.

PubMed

Jiangrong Shen; Kang Lin; Yueming Wang; Gang Pan

2017-07-01

Spiking neural networks are biologically plausible and power-efficient on neuromorphic hardware, while recurrent neural networks have been proven to be efficient on time series data. However, how to use the recurrent property to improve the performance of spiking neural networks is still a problem. This paper proposes a recurrent spiking neural network for character recognition using trajectories. In the network, a new encoding method is designed, in which varying time ranges of input streams are used in different recurrent layers. This is able to improve the generalization ability of our model compared with general encoding methods. The experiments are conducted on four groups of the character data set from University of Edinburgh. The results show that our method can achieve a higher average recognition accuracy than existing methods.

Design of a universal two-layered neural network derived from the PLI theory

NASA Astrophysics Data System (ADS)

Hu, Chia-Lun J.

2004-05-01

The if-and-only-if (IFF) condition that a set of M analog-to-digital vector-mapping relations can be learned by a one-layered-feed-forward neural network (OLNN) is that all the input analog vectors dichotomized by the i-th output bit must be positively, linearly independent, or PLI. If they are not PLI, then the OLNN just cannot learn no matter what learning rules is employed because the solution of the connection matrix does not exist mathematically. However, in this case, one can still design a parallel-cascaded, two-layered, perceptron (PCTLP) to acheive this general mapping goal. The design principle of this "universal" neural network is derived from the major mathematical properties of the PLI theory - changing the output bits of the dependent relations existing among the dichotomized input vectors to make the PLD relations PLI. Then with a vector concatenation technique, the required mapping can still be learned by this PCTLP system with very high efficiency. This paper will report in detail the mathematical derivation of the general design principle and the design procedures of the PCTLP neural network system. It then will be verified in general by a practical numerical example.

Coronary Artery Diagnosis Aided by Neural Network

NASA Astrophysics Data System (ADS)

Stefko, Kamil

2007-01-01

Coronary artery disease is due to atheromatous narrowing and subsequent occlusion of the coronary vessel. Application of optimised feed forward multi-layer back propagation neural network (MLBP) for detection of narrowing in coronary artery vessels is presented in this paper. The research was performed using 580 data records from traditional ECG exercise test confirmed by coronary arteriography results. Each record of training database included description of the state of a patient providing input data for the neural network. Level and slope of ST segment of a 12 lead ECG signal recorded at rest and after effort (48 floating point values) was the main component of input data for neural network was. Coronary arteriography results (verified the existence or absence of more than 50% stenosis of the particular coronary vessels) were used as a correct neural network training output pattern. More than 96% of cases were correctly recognised by especially optimised and a thoroughly verified neural network. Leave one out method was used for neural network verification so 580 data records could be used for training as well as for verification of neural network.

Finite-time convergent recurrent neural network with a hard-limiting activation function for constrained optimization with piecewise-linear objective functions.

PubMed

Liu, Qingshan; Wang, Jun

2011-04-01

This paper presents a one-layer recurrent neural network for solving a class of constrained nonsmooth optimization problems with piecewise-linear objective functions. The proposed neural network is guaranteed to be globally convergent in finite time to the optimal solutions under a mild condition on a derived lower bound of a single gain parameter in the model. The number of neurons in the neural network is the same as the number of decision variables of the optimization problem. Compared with existing neural networks for optimization, the proposed neural network has a couple of salient features such as finite-time convergence and a low model complexity. Specific models for two important special cases, namely, linear programming and nonsmooth optimization, are also presented. In addition, applications to the shortest path problem and constrained least absolute deviation problem are discussed with simulation results to demonstrate the effectiveness and characteristics of the proposed neural network.

Deep architecture neural network-based real-time image processing for image-guided radiotherapy.

PubMed

Mori, Shinichiro

2017-08-01

To develop real-time image processing for image-guided radiotherapy, we evaluated several neural network models for use with different imaging modalities, including X-ray fluoroscopic image denoising. Setup images of prostate cancer patients were acquired with two oblique X-ray fluoroscopic units. Two types of residual network were designed: a convolutional autoencoder (rCAE) and a convolutional neural network (rCNN). We changed the convolutional kernel size and number of convolutional layers for both networks, and the number of pooling and upsampling layers for rCAE. The ground-truth image was applied to the contrast-limited adaptive histogram equalization (CLAHE) method of image processing. Network models were trained to keep the quality of the output image close to that of the ground-truth image from the input image without image processing. For image denoising evaluation, noisy input images were used for the training. More than 6 convolutional layers with convolutional kernels >5×5 improved image quality. However, this did not allow real-time imaging. After applying a pair of pooling and upsampling layers to both networks, rCAEs with >3 convolutions each and rCNNs with >12 convolutions with a pair of pooling and upsampling layers achieved real-time processing at 30 frames per second (fps) with acceptable image quality. Use of our suggested network achieved real-time image processing for contrast enhancement and image denoising by the use of a conventional modern personal computer. Copyright © 2017 Associazione Italiana di Fisica Medica. Published by Elsevier Ltd. All rights reserved.

Reynolds averaged turbulence modelling using deep neural networks with embedded invariance

DOE PAGES

Ling, Julia; Kurzawski, Andrew; Templeton, Jeremy

2016-10-18

There exists significant demand for improved Reynolds-averaged Navier–Stokes (RANS) turbulence models that are informed by and can represent a richer set of turbulence physics. This paper presents a method of using deep neural networks to learn a model for the Reynolds stress anisotropy tensor from high-fidelity simulation data. A novel neural network architecture is proposed which uses a multiplicative layer with an invariant tensor basis to embed Galilean invariance into the predicted anisotropy tensor. It is demonstrated that this neural network architecture provides improved prediction accuracy compared with a generic neural network architecture that does not embed this invariance property.more » Furthermore, the Reynolds stress anisotropy predictions of this invariant neural network are propagated through to the velocity field for two test cases. For both test cases, significant improvement versus baseline RANS linear eddy viscosity and nonlinear eddy viscosity models is demonstrated.« less

Reynolds averaged turbulence modelling using deep neural networks with embedded invariance

DOE Office of Scientific and Technical Information (OSTI.GOV)

Ling, Julia; Kurzawski, Andrew; Templeton, Jeremy

There exists significant demand for improved Reynolds-averaged Navier–Stokes (RANS) turbulence models that are informed by and can represent a richer set of turbulence physics. This paper presents a method of using deep neural networks to learn a model for the Reynolds stress anisotropy tensor from high-fidelity simulation data. A novel neural network architecture is proposed which uses a multiplicative layer with an invariant tensor basis to embed Galilean invariance into the predicted anisotropy tensor. It is demonstrated that this neural network architecture provides improved prediction accuracy compared with a generic neural network architecture that does not embed this invariance property.more » Furthermore, the Reynolds stress anisotropy predictions of this invariant neural network are propagated through to the velocity field for two test cases. For both test cases, significant improvement versus baseline RANS linear eddy viscosity and nonlinear eddy viscosity models is demonstrated.« less

Resolution of Singularities Introduced by Hierarchical Structure in Deep Neural Networks.

PubMed

Nitta, Tohru

2017-10-01

We present a theoretical analysis of singular points of artificial deep neural networks, resulting in providing deep neural network models having no critical points introduced by a hierarchical structure. It is considered that such deep neural network models have good nature for gradient-based optimization. First, we show that there exist a large number of critical points introduced by a hierarchical structure in deep neural networks as straight lines, depending on the number of hidden layers and the number of hidden neurons. Second, we derive a sufficient condition for deep neural networks having no critical points introduced by a hierarchical structure, which can be applied to general deep neural networks. It is also shown that the existence of critical points introduced by a hierarchical structure is determined by the rank and the regularity of weight matrices for a specific class of deep neural networks. Finally, two kinds of implementation methods of the sufficient conditions to have no critical points are provided. One is a learning algorithm that can avoid critical points introduced by the hierarchical structure during learning (called avoidant learning algorithm). The other is a neural network that does not have some critical points introduced by the hierarchical structure as an inherent property (called avoidant neural network).

An alternative approach for neural network evolution with a genetic algorithm: crossover by combinatorial optimization.

PubMed

García-Pedrajas, Nicolás; Ortiz-Boyer, Domingo; Hervás-Martínez, César

2006-05-01

In this work we present a new approach to crossover operator in the genetic evolution of neural networks. The most widely used evolutionary computation paradigm for neural network evolution is evolutionary programming. This paradigm is usually preferred due to the problems caused by the application of crossover to neural network evolution. However, crossover is the most innovative operator within the field of evolutionary computation. One of the most notorious problems with the application of crossover to neural networks is known as the permutation problem. This problem occurs due to the fact that the same network can be represented in a genetic coding by many different codifications. Our approach modifies the standard crossover operator taking into account the special features of the individuals to be mated. We present a new model for mating individuals that considers the structure of the hidden layer and redefines the crossover operator. As each hidden node represents a non-linear projection of the input variables, we approach the crossover as a problem on combinatorial optimization. We can formulate the problem as the extraction of a subset of near-optimal projections to create the hidden layer of the new network. This new approach is compared to a classical crossover in 25 real-world problems with an excellent performance. Moreover, the networks obtained are much smaller than those obtained with classical crossover operator.

Function approximation using combined unsupervised and supervised learning.

PubMed

Andras, Peter

2014-03-01

Function approximation is one of the core tasks that are solved using neural networks in the context of many engineering problems. However, good approximation results need good sampling of the data space, which usually requires exponentially increasing volume of data as the dimensionality of the data increases. At the same time, often the high-dimensional data is arranged around a much lower dimensional manifold. Here we propose the breaking of the function approximation task for high-dimensional data into two steps: (1) the mapping of the high-dimensional data onto a lower dimensional space corresponding to the manifold on which the data resides and (2) the approximation of the function using the mapped lower dimensional data. We use over-complete self-organizing maps (SOMs) for the mapping through unsupervised learning, and single hidden layer neural networks for the function approximation through supervised learning. We also extend the two-step procedure by considering support vector machines and Bayesian SOMs for the determination of the best parameters for the nonlinear neurons in the hidden layer of the neural networks used for the function approximation. We compare the approximation performance of the proposed neural networks using a set of functions and show that indeed the neural networks using combined unsupervised and supervised learning outperform in most cases the neural networks that learn the function approximation using the original high-dimensional data.

Human Age Recognition by Electrocardiogram Signal Based on Artificial Neural Network

NASA Astrophysics Data System (ADS)

Dasgupta, Hirak

2016-12-01

The objective of this work is to make a neural network function approximation model to detect human age from the electrocardiogram (ECG) signal. The input vectors of the neural network are the Katz fractal dimension of the ECG signal, frequencies in the QRS complex, male or female (represented by numeric constant) and the average of successive R-R peak distance of a particular ECG signal. The QRS complex has been detected by short time Fourier transform algorithm. The successive R peak has been detected by, first cutting the signal into periods by auto-correlation method and then finding the absolute of the highest point in each period. The neural network used in this problem consists of two layers, with Sigmoid neuron in the input and linear neuron in the output layer. The result shows the mean of errors as -0.49, 1.03, 0.79 years and the standard deviation of errors as 1.81, 1.77, 2.70 years during training, cross validation and testing with unknown data sets, respectively.

Method Accelerates Training Of Some Neural Networks

NASA Technical Reports Server (NTRS)

Shelton, Robert O.

1992-01-01

Three-layer networks trained faster provided two conditions are satisfied: numbers of neurons in layers are such that majority of work done in synaptic connections between input and hidden layers, and number of neurons in input layer at least as great as number of training pairs of input and output vectors. Based on modified version of back-propagation method.

Unfolding the neutron spectrum of a NE213 scintillator using artificial neural networks.

PubMed

Sharghi Ido, A; Bonyadi, M R; Etaati, G R; Shahriari, M

2009-10-01

Artificial neural networks technology has been applied to unfold the neutron spectra from the pulse height distribution measured with NE213 liquid scintillator. Here, both the single and multi-layer perceptron neural network models have been implemented to unfold the neutron spectrum from an Am-Be neutron source. The activation function and the connectivity of the neurons have been investigated and the results have been analyzed in terms of the network's performance. The simulation results show that the neural network that utilizes the Satlins transfer function has the best performance. In addition, omitting the bias connection of the neurons improve the performance of the network. Also, the SCINFUL code is used for generating the response functions in the training phase of the process. Finally, the results of the neural network simulation have been compared with those of the FORIST unfolding code for both (241)Am-Be and (252)Cf neutron sources. The results of neural network are in good agreement with FORIST code.

A neural-network-based model for the dynamic simulation of the tire/suspension system while traversing road irregularities.

PubMed

Guarneri, Paolo; Rocca, Gianpiero; Gobbi, Massimiliano

2008-09-01

This paper deals with the simulation of the tire/suspension dynamics by using recurrent neural networks (RNNs). RNNs are derived from the multilayer feedforward neural networks, by adding feedback connections between output and input layers. The optimal network architecture derives from a parametric analysis based on the optimal tradeoff between network accuracy and size. The neural network can be trained with experimental data obtained in the laboratory from simulated road profiles (cleats). The results obtained from the neural network demonstrate good agreement with the experimental results over a wide range of operation conditions. The NN model can be effectively applied as a part of vehicle system model to accurately predict elastic bushings and tire dynamics behavior. Although the neural network model, as a black-box model, does not provide a good insight of the physical behavior of the tire/suspension system, it is a useful tool for assessing vehicle ride and noise, vibration, harshness (NVH) performance due to its good computational efficiency and accuracy.

Prediction of Industrial Electric Energy Consumption in Anhui Province Based on GA-BP Neural Network

NASA Astrophysics Data System (ADS)

Zhang, Jiajing; Yin, Guodong; Ni, Youcong; Chen, Jinlan

2018-01-01

In order to improve the prediction accuracy of industrial electrical energy consumption, a prediction model of industrial electrical energy consumption was proposed based on genetic algorithm and neural network. The model use genetic algorithm to optimize the weights and thresholds of BP neural network, and the model is used to predict the energy consumption of industrial power in Anhui Province, to improve the prediction accuracy of industrial electric energy consumption in Anhui province. By comparing experiment of GA-BP prediction model and BP neural network model, the GA-BP model is more accurate with smaller number of neurons in the hidden layer.

The optimization of force inputs for active structural acoustic control using a neural network

NASA Technical Reports Server (NTRS)

Cabell, R. H.; Lester, H. C.; Silcox, R. J.

1992-01-01

This paper investigates the use of a neural network to determine which force actuators, of a multi-actuator array, are best activated in order to achieve structural-acoustic control. The concept is demonstrated using a cylinder/cavity model on which the control forces, produced by piezoelectric actuators, are applied with the objective of reducing the interior noise. A two-layer neural network is employed and the back propagation solution is compared with the results calculated by a conventional, least-squares optimization analysis. The ability of the neural network to accurately and efficiently control actuator activation for interior noise reduction is demonstrated.

Comparison between extreme learning machine and wavelet neural networks in data classification

NASA Astrophysics Data System (ADS)

Yahia, Siwar; Said, Salwa; Jemai, Olfa; Zaied, Mourad; Ben Amar, Chokri

2017-03-01

Extreme learning Machine is a well known learning algorithm in the field of machine learning. It's about a feed forward neural network with a single-hidden layer. It is an extremely fast learning algorithm with good generalization performance. In this paper, we aim to compare the Extreme learning Machine with wavelet neural networks, which is a very used algorithm. We have used six benchmark data sets to evaluate each technique. These datasets Including Wisconsin Breast Cancer, Glass Identification, Ionosphere, Pima Indians Diabetes, Wine Recognition and Iris Plant. Experimental results have shown that both extreme learning machine and wavelet neural networks have reached good results.

The use of neural network technology to model swimming performance.

PubMed

Silva, António José; Costa, Aldo Manuel; Oliveira, Paulo Moura; Reis, Victor Machado; Saavedra, José; Perl, Jurgen; Rouboa, Abel; Marinho, Daniel Almeida

2007-01-01

to identify the factors which are able to explain the performance in the 200 meters individual medley and 400 meters front crawl events in young swimmers, to model the performance in those events using non-linear mathematic methods through artificial neural networks (multi-layer perceptrons) and to assess the neural network models precision to predict the performance. A sample of 138 young swimmers (65 males and 73 females) of national level was submitted to a test battery comprising four different domains: kinanthropometric evaluation, dry land functional evaluation (strength and flexibility), swimming functional evaluation (hydrodynamics, hydrostatic and bioenergetics characteristics) and swimming technique evaluation. To establish a profile of the young swimmer non-linear combinations between preponderant variables for each gender and swim performance in the 200 meters medley and 400 meters font crawl events were developed. For this purpose a feed forward neural network was used (Multilayer Perceptron) with three neurons in a single hidden layer. The prognosis precision of the model (error lower than 0.8% between true and estimated performances) is supported by recent evidence. Therefore, we consider that the neural network tool can be a good approach in the resolution of complex problems such as performance modeling and the talent identification in swimming and, possibly, in a wide variety of sports. Key pointsThe non-linear analysis resulting from the use of feed forward neural network allowed us the development of four performance models.The mean difference between the true and estimated results performed by each one of the four neural network models constructed was low.The neural network tool can be a good approach in the resolution of the performance modeling as an alternative to the standard statistical models that presume well-defined distributions and independence among all inputs.The use of neural networks for sports sciences application allowed us to create very realistic models for swimming performance prediction based on previous selected criterions that were related with the dependent variable (performance).

Anomalous Signal Detection in ELF Band Electromagnetic Wave using Multi-layer Neural Network with Wavelet Decomposition

NASA Astrophysics Data System (ADS)

Itai, Akitoshi; Yasukawa, Hiroshi; Takumi, Ichi; Hata, Masayasu

It is well known that electromagnetic waves radiated from the earth's crust are useful for predicting earthquakes. We analyze the electromagnetic waves received at the extremely low frequency band of 223Hz. These observed signals contain the seismic radiation from the earth's crust, but also include several undesired signals. Our research focuses on the signal detection technique to identify an anomalous signal corresponding to the seismic radiation in the observed signal. Conventional anomalous signal detections lack a wide applicability due to their assumptions, e.g. the digital data have to be observed at the same time or the same sensor. In order to overcome the limitation related to the observed signal, we proposed the anomalous signals detection based on a multi-layer neural network which is trained by digital data observed during a span of a day. In the neural network approach, training data do not need to be recorded at the same place or the same time. However, some noises, which have a large amplitude, are detected as the anomalous signal. This paper develops a multi-layer neural network to decrease the false detection of the anomalous signal from the electromagnetic wave. The training data for the proposed network is the decomposed signal of the observed signal during several days, since the seismic radiations are often recorded from several days to a couple of weeks. Results show that the proposed neural network is useful to achieve the accurate detection of the anomalous signal that indicates seismic activity.

Classification of remotely sensed data using OCR-inspired neural network techniques. [Optical Character Recognition

NASA Technical Reports Server (NTRS)

Kiang, Richard K.

1992-01-01

Neural networks have been applied to classifications of remotely sensed data with some success. To improve the performance of this approach, an examination was made of how neural networks are applied to the optical character recognition (OCR) of handwritten digits and letters. A three-layer, feedforward network, along with techniques adopted from OCR, was used to classify Landsat-4 Thematic Mapper data. Good results were obtained. To overcome the difficulties that are characteristic of remote sensing applications and to attain significant improvements in classification accuracy, a special network architecture may be required.

Neural synchronization as a hypothetical explanation of the psychoanalytic unconscious.

PubMed

Ceylan, Mehmet Emin; Dönmez, Aslıhan; Ünsalver, Barış Önen; Evrensel, Alper

2016-02-01

Cognitive scientists have tried to explain the neural mechanisms of unconscious mental states such as coma, epileptic seizures, and anesthesia-induced unconsciousness. However these types of unconscious states are different from the psychoanalytic unconscious. In this review, we aim to present our hypothesis about the neural correlates underlying psychoanalytic unconscious. To fulfill this aim, we firstly review the previous explanations about the neural correlates of conscious and unconscious mental states, such as brain oscillations, synchronicity of neural networks, and cognitive binding. By doing so, we hope to lay a neuroscientific ground for our hypothesis about neural correlates of psychoanalytic unconscious; parallel but unsynchronized neural networks between different layers of consciousness and unconsciousness. Next, we propose a neuroscientific mechanism about how the repressed mental events reach the conscious awareness; the lock of neural synchronization between two mental layers of conscious and unconscious. At the last section, we will discuss the data about schizophrenia as a clinical example of our proposed hypothesis. Copyright © 2015 Elsevier Inc. All rights reserved.

Financial Time Series Prediction Using Spiking Neural Networks

PubMed Central

Reid, David; Hussain, Abir Jaafar; Tawfik, Hissam

2014-01-01

In this paper a novel application of a particular type of spiking neural network, a Polychronous Spiking Network, was used for financial time series prediction. It is argued that the inherent temporal capabilities of this type of network are suited to non-stationary data such as this. The performance of the spiking neural network was benchmarked against three systems: two “traditional”, rate-encoded, neural networks; a Multi-Layer Perceptron neural network and a Dynamic Ridge Polynomial neural network, and a standard Linear Predictor Coefficients model. For this comparison three non-stationary and noisy time series were used: IBM stock data; US/Euro exchange rate data, and the price of Brent crude oil. The experiments demonstrated favourable prediction results for the Spiking Neural Network in terms of Annualised Return and prediction error for 5-Step ahead predictions. These results were also supported by other relevant metrics such as Maximum Drawdown and Signal-To-Noise ratio. This work demonstrated the applicability of the Polychronous Spiking Network to financial data forecasting and this in turn indicates the potential of using such networks over traditional systems in difficult to manage non-stationary environments. PMID:25170618

A training rule which guarantees finite-region stability for a class of closed-loop neural-network control systems.

PubMed

Kuntanapreeda, S; Fullmer, R R

1996-01-01

A training method for a class of neural network controllers is presented which guarantees closed-loop system stability. The controllers are assumed to be nonlinear, feedforward, sampled-data, full-state regulators implemented as single hidden-layer neural networks. The controlled systems must be locally hermitian and observable. Stability of the closed-loop system is demonstrated by determining a Lyapunov function, which can be used to identify a finite stability region about the regulator point.

Design of double fuzzy clustering-driven context neural networks.

PubMed

Kim, Eun-Hu; Oh, Sung-Kwun; Pedrycz, Witold

2018-08-01

In this study, we introduce a novel category of double fuzzy clustering-driven context neural networks (DFCCNNs). The study is focused on the development of advanced design methodologies for redesigning the structure of conventional fuzzy clustering-based neural networks. The conventional fuzzy clustering-based neural networks typically focus on dividing the input space into several local spaces (implied by clusters). In contrast, the proposed DFCCNNs take into account two distinct local spaces called context and cluster spaces, respectively. Cluster space refers to the local space positioned in the input space whereas context space concerns a local space formed in the output space. Through partitioning the output space into several local spaces, each context space is used as the desired (target) local output to construct local models. To complete this, the proposed network includes a new context layer for reasoning about context space in the output space. In this sense, Fuzzy C-Means (FCM) clustering is useful to form local spaces in both input and output spaces. The first one is used in order to form clusters and train weights positioned between the input and hidden layer, whereas the other one is applied to the output space to form context spaces. The key features of the proposed DFCCNNs can be enumerated as follows: (i) the parameters between the input layer and hidden layer are built through FCM clustering. The connections (weights) are specified as constant terms being in fact the centers of the clusters. The membership functions (represented through the partition matrix) produced by the FCM are used as activation functions located at the hidden layer of the "conventional" neural networks. (ii) Following the hidden layer, a context layer is formed to approximate the context space of the output variable and each node in context layer means individual local model. The outputs of the context layer are specified as a combination of both weights formed as linear function and the outputs of the hidden layer. The weights are updated using the least square estimation (LSE)-based method. (iii) At the output layer, the outputs of context layer are decoded to produce the corresponding numeric output. At this time, the weighted average is used and the weights are also adjusted with the use of the LSE scheme. From the viewpoint of performance improvement, the proposed design methodologies are discussed and experimented with the aid of benchmark machine learning datasets. Through the experiments, it is shown that the generalization abilities of the proposed DFCCNNs are better than those of the conventional FCNNs reported in the literature. Copyright © 2018 Elsevier Ltd. All rights reserved.

Neural networks within multi-core optic fibers

PubMed Central

Cohen, Eyal; Malka, Dror; Shemer, Amir; Shahmoon, Asaf; Zalevsky, Zeev; London, Michael

2016-01-01

Hardware implementation of artificial neural networks facilitates real-time parallel processing of massive data sets. Optical neural networks offer low-volume 3D connectivity together with large bandwidth and minimal heat production in contrast to electronic implementation. Here, we present a conceptual design for in-fiber optical neural networks. Neurons and synapses are realized as individual silica cores in a multi-core fiber. Optical signals are transferred transversely between cores by means of optical coupling. Pump driven amplification in erbium-doped cores mimics synaptic interactions. We simulated three-layered feed-forward neural networks and explored their capabilities. Simulations suggest that networks can differentiate between given inputs depending on specific configurations of amplification; this implies classification and learning capabilities. Finally, we tested experimentally our basic neuronal elements using fibers, couplers, and amplifiers, and demonstrated that this configuration implements a neuron-like function. Therefore, devices similar to our proposed multi-core fiber could potentially serve as building blocks for future large-scale small-volume optical artificial neural networks. PMID:27383911

Neural networks within multi-core optic fibers.

PubMed

Cohen, Eyal; Malka, Dror; Shemer, Amir; Shahmoon, Asaf; Zalevsky, Zeev; London, Michael

2016-07-07

Hardware implementation of artificial neural networks facilitates real-time parallel processing of massive data sets. Optical neural networks offer low-volume 3D connectivity together with large bandwidth and minimal heat production in contrast to electronic implementation. Here, we present a conceptual design for in-fiber optical neural networks. Neurons and synapses are realized as individual silica cores in a multi-core fiber. Optical signals are transferred transversely between cores by means of optical coupling. Pump driven amplification in erbium-doped cores mimics synaptic interactions. We simulated three-layered feed-forward neural networks and explored their capabilities. Simulations suggest that networks can differentiate between given inputs depending on specific configurations of amplification; this implies classification and learning capabilities. Finally, we tested experimentally our basic neuronal elements using fibers, couplers, and amplifiers, and demonstrated that this configuration implements a neuron-like function. Therefore, devices similar to our proposed multi-core fiber could potentially serve as building blocks for future large-scale small-volume optical artificial neural networks.

Application of artificial neural networks to chemostratigraphy

NASA Astrophysics Data System (ADS)

Malmgren, BjöRn A.; Nordlund, Ulf

1996-08-01

Artificial neural networks, a branch of artificial intelligence, are computer systems formed by a number of simple, highly interconnected processing units that have the ability to learn a set of target vectors from a set of associated input signals. Neural networks learn by self-adjusting a set of parameters, using some pertinent algorithm to minimize the error between the desired output and network output. We explore the potential of this approach in solving a problem involving classification of geochemical data. The data, taken from the literature, are derived from four late Quaternary zones of volcanic ash of basaltic and rhyolithic origin from the Norwegian Sea. These ash layers span the oxygen isotope zones 1, 5, 7, and 11, respectively (last 420,000 years). The data consist of nine geochemical variables (oxides) determined in each of 183 samples. We employed a three-layer back propagation neural network to assess its efficiency to optimally differentiate samples from the four ash zones on the basis of their geochemical composition. For comparison, three statistical pattern recognition techniques, linear discriminant analysis, the k-nearest neighbor (k-NN) technique, and SIMCA (soft independent modeling of class analogy), were applied to the same data. All of these showed considerably higher error rates than the artificial neural network, indicating that the back propagation network was indeed more powerful in correctly classifying the ash particles to the appropriate zone on the basis of their geochemical composition.

Extraction of texture features with a multiresolution neural network

NASA Astrophysics Data System (ADS)

Lepage, Richard; Laurendeau, Denis; Gagnon, Roger A.

1992-09-01

Texture is an important surface characteristic. Many industrial materials such as wood, textile, or paper are best characterized by their texture. Detection of defaults occurring on such materials or classification for quality control anD matching can be carried out through careful texture analysis. A system for the classification of pieces of wood used in the furniture industry is proposed. This paper is concerned with a neural network implementation of the features extraction and classification components of the proposed system. Texture appears differently depending at which spatial scale it is observed. A complete description of a texture thus implies an analysis at several spatial scales. We propose a compact pyramidal representation of the input image for multiresolution analysis. The feature extraction system is implemented on a multilayer artificial neural network. Each level of the pyramid, which is a representation of the input image at a given spatial resolution scale, is mapped into a layer of the neural network. A full resolution texture image is input at the base of the pyramid and a representation of the texture image at multiple resolutions is generated by the feedforward pyramid structure of the neural network. The receptive field of each neuron at a given pyramid level is preprogrammed as a discrete Gaussian low-pass filter. Meaningful characteristics of the textured image must be extracted if a good resolving power of the classifier must be achieved. Local dominant orientation is the principal feature which is extracted from the textured image. Local edge orientation is computed with a Sobel mask at four orientation angles (multiple of (pi) /4). The resulting intrinsic image, that is, the local dominant orientation image, is fed to the texture classification neural network. The classification network is a three-layer feedforward back-propagation neural network.

Predicting Item Difficulty in a Reading Comprehension Test with an Artificial Neural Network.

ERIC Educational Resources Information Center

Perkins, Kyle; And Others

This paper reports the results of using a three-layer backpropagation artificial neural network to predict item difficulty in a reading comprehension test. Two network structures were developed, one with and one without a sigmoid function in the output processing unit. The data set, which consisted of a table of coded test items and corresponding…

Optimizing hidden layer node number of BP network to estimate fetal weight

NASA Astrophysics Data System (ADS)

Su, Juan; Zou, Yuanwen; Lin, Jiangli; Wang, Tianfu; Li, Deyu; Xie, Tao

2007-12-01

The ultrasonic estimation of fetal weigh before delivery is of most significance for obstetrical clinic. Estimating fetal weight more accurately is crucial for prenatal care, obstetrical treatment, choosing appropriate delivery methods, monitoring fetal growth and reducing the risk of newborn complications. In this paper, we introduce a method which combines golden section and artificial neural network (ANN) to estimate the fetal weight. The golden section is employed to optimize the hidden layer node number of the back propagation (BP) neural network. The method greatly improves the accuracy of fetal weight estimation, and simultaneously avoids choosing the hidden layer node number with subjective experience. The estimation coincidence rate achieves 74.19%, and the mean absolute error is 185.83g.

The algorithm study for using the back propagation neural network in CT image segmentation

NASA Astrophysics Data System (ADS)

Zhang, Peng; Liu, Jie; Chen, Chen; Li, Ying Qi

2017-01-01

Back propagation neural network(BP neural network) is a type of multi-layer feed forward network which spread positively, while the error spread backwardly. Since BP network has advantages in learning and storing the mapping between a large number of input and output layers without complex mathematical equations to describe the mapping relationship, it is most widely used. BP can iteratively compute the weight coefficients and thresholds of the network based on the training and back propagation of samples, which can minimize the error sum of squares of the network. Since the boundary of the computed tomography (CT) heart images is usually discontinuous, and it exist large changes in the volume and boundary of heart images, The conventional segmentation such as region growing and watershed algorithm can't achieve satisfactory results. Meanwhile, there are large differences between the diastolic and systolic images. The conventional methods can't accurately classify the two cases. In this paper, we introduced BP to handle the segmentation of heart images. We segmented a large amount of CT images artificially to obtain the samples, and the BP network was trained based on these samples. To acquire the appropriate BP network for the segmentation of heart images, we normalized the heart images, and extract the gray-level information of the heart. Then the boundary of the images was input into the network to compare the differences between the theoretical output and the actual output, and we reinput the errors into the BP network to modify the weight coefficients of layers. Through a large amount of training, the BP network tend to be stable, and the weight coefficients of layers can be determined, which means the relationship between the CT images and the boundary of heart.

Using Upper Layer Weights to Efficiently Construct and Train Feedforward Neural Networks Executing Backpropagation

DTIC Science & Technology

2011-03-01

algorithm is utilized by Belue, Steppe, & Bauer and Kocur , et al. (Belue, Steppe, & Bauer, April 1996) ( Kocur , et al., 1996). Bacauskiene and...Society. Cardiff, UK. Kocur , C., Roger, S., Myers, L., Burns, T., Hoffmeister, J., Bauer, K., et al. (1996). Using neural networks to select

Neural Networks

NASA Astrophysics Data System (ADS)

Schwindling, Jerome

2010-04-01

This course presents an overview of the concepts of the neural networks and their aplication in the framework of High energy physics analyses. After a brief introduction on the concept of neural networks, the concept is explained in the frame of neuro-biology, introducing the concept of multi-layer perceptron, learning and their use as data classifer. The concept is then presented in a second part using in more details the mathematical approach focussing on typical use cases faced in particle physics. Finally, the last part presents the best way to use such statistical tools in view of event classifers, putting the emphasis on the setup of the multi-layer perceptron. The full article (15 p.) corresponding to this lecture is written in french and is provided in the proceedings of the book SOS 2008.

Adaptive nonlinear polynomial neural networks for control of boundary layer/structural interaction

NASA Technical Reports Server (NTRS)

Parker, B. Eugene, Jr.; Cellucci, Richard L.; Abbott, Dean W.; Barron, Roger L.; Jordan, Paul R., III; Poor, H. Vincent

1993-01-01

The acoustic pressures developed in a boundary layer can interact with an aircraft panel to induce significant vibration in the panel. Such vibration is undesirable due to the aerodynamic drag and structure-borne cabin noises that result. The overall objective of this work is to develop effective and practical feedback control strategies for actively reducing this flow-induced structural vibration. This report describes the results of initial evaluations using polynomial, neural network-based, feedback control to reduce flow induced vibration in aircraft panels due to turbulent boundary layer/structural interaction. Computer simulations are used to develop and analyze feedback control strategies to reduce vibration in a beam as a first step. The key differences between this work and that going on elsewhere are as follows: that turbulent and transitional boundary layers represent broadband excitation and thus present a more complex stochastic control scenario than that of narrow band (e.g., laminar boundary layer) excitation; and secondly, that the proposed controller structures are adaptive nonlinear infinite impulse response (IIR) polynomial neural network, as opposed to the traditional adaptive linear finite impulse response (FIR) filters used in most studies to date. The controllers implemented in this study achieved vibration attenuation of 27 to 60 dB depending on the type of boundary layer established by laminar, turbulent, and intermittent laminar-to-turbulent transitional flows. Application of multi-input, multi-output, adaptive, nonlinear feedback control of vibration in aircraft panels based on polynomial neural networks appears to be feasible today. Plans are outlined for Phase 2 of this study, which will include extending the theoretical investigation conducted in Phase 2 and verifying the results in a series of laboratory experiments involving both bum and plate models.

A neural network model of ventriloquism effect and aftereffect.

PubMed

Magosso, Elisa; Cuppini, Cristiano; Ursino, Mauro

2012-01-01

Presenting simultaneous but spatially discrepant visual and auditory stimuli induces a perceptual translocation of the sound towards the visual input, the ventriloquism effect. General explanation is that vision tends to dominate over audition because of its higher spatial reliability. The underlying neural mechanisms remain unclear. We address this question via a biologically inspired neural network. The model contains two layers of unimodal visual and auditory neurons, with visual neurons having higher spatial resolution than auditory ones. Neurons within each layer communicate via lateral intra-layer synapses; neurons across layers are connected via inter-layer connections. The network accounts for the ventriloquism effect, ascribing it to a positive feedback between the visual and auditory neurons, triggered by residual auditory activity at the position of the visual stimulus. Main results are: i) the less localized stimulus is strongly biased toward the most localized stimulus and not vice versa; ii) amount of the ventriloquism effect changes with visual-auditory spatial disparity; iii) ventriloquism is a robust behavior of the network with respect to parameter value changes. Moreover, the model implements Hebbian rules for potentiation and depression of lateral synapses, to explain ventriloquism aftereffect (that is, the enduring sound shift after exposure to spatially disparate audio-visual stimuli). By adaptively changing the weights of lateral synapses during cross-modal stimulation, the model produces post-adaptive shifts of auditory localization that agree with in-vivo observations. The model demonstrates that two unimodal layers reciprocally interconnected may explain ventriloquism effect and aftereffect, even without the presence of any convergent multimodal area. The proposed study may provide advancement in understanding neural architecture and mechanisms at the basis of visual-auditory integration in the spatial realm.

Experimental fault characterization of a neural network

NASA Technical Reports Server (NTRS)

Tan, Chang-Huong

1990-01-01

The effects of a variety of faults on a neural network is quantified via simulation. The neural network consists of a single-layered clustering network and a three-layered classification network. The percentage of vectors mistagged by the clustering network, the percentage of vectors misclassified by the classification network, the time taken for the network to stabilize, and the output values are all measured. The results show that both transient and permanent faults have a significant impact on the performance of the measured network. The corresponding mistag and misclassification percentages are typically within 5 to 10 percent of each other. The average mistag percentage and the average misclassification percentage are both about 25 percent. After relearning, the percentage of misclassifications is reduced to 9 percent. In addition, transient faults are found to cause the network to be increasingly unstable as the duration of a transient is increased. The impact of link faults is relatively insignificant in comparison with node faults (1 versus 19 percent misclassified after relearning). There is a linear increase in the mistag and misclassification percentages with decreasing hardware redundancy. In addition, the mistag and misclassification percentages linearly decrease with increasing network size.

Flank wears Simulation by using back propagation neural network when cutting hardened H-13 steel in CNC End Milling

NASA Astrophysics Data System (ADS)

Hazza, Muataz Hazza F. Al; Adesta, Erry Y. T.; Riza, Muhammad

2013-12-01

High speed milling has many advantages such as higher removal rate and high productivity. However, higher cutting speed increase the flank wear rate and thus reducing the cutting tool life. Therefore estimating and predicting the flank wear length in early stages reduces the risk of unaccepted tooling cost. This research presents a neural network model for predicting and simulating the flank wear in the CNC end milling process. A set of sparse experimental data for finish end milling on AISI H13 at hardness of 48 HRC have been conducted to measure the flank wear length. Then the measured data have been used to train the developed neural network model. Artificial neural network (ANN) was applied to predict the flank wear length. The neural network contains twenty hidden layer with feed forward back propagation hierarchical. The neural network has been designed with MATLAB Neural Network Toolbox. The results show a high correlation between the predicted and the observed flank wear which indicates the validity of the models.

Phylogenetic convolutional neural networks in metagenomics.

PubMed

Fioravanti, Diego; Giarratano, Ylenia; Maggio, Valerio; Agostinelli, Claudio; Chierici, Marco; Jurman, Giuseppe; Furlanello, Cesare

2018-03-08

Convolutional Neural Networks can be effectively used only when data are endowed with an intrinsic concept of neighbourhood in the input space, as is the case of pixels in images. We introduce here Ph-CNN, a novel deep learning architecture for the classification of metagenomics data based on the Convolutional Neural Networks, with the patristic distance defined on the phylogenetic tree being used as the proximity measure. The patristic distance between variables is used together with a sparsified version of MultiDimensional Scaling to embed the phylogenetic tree in a Euclidean space. Ph-CNN is tested with a domain adaptation approach on synthetic data and on a metagenomics collection of gut microbiota of 38 healthy subjects and 222 Inflammatory Bowel Disease patients, divided in 6 subclasses. Classification performance is promising when compared to classical algorithms like Support Vector Machines and Random Forest and a baseline fully connected neural network, e.g. the Multi-Layer Perceptron. Ph-CNN represents a novel deep learning approach for the classification of metagenomics data. Operatively, the algorithm has been implemented as a custom Keras layer taking care of passing to the following convolutional layer not only the data but also the ranked list of neighbourhood of each sample, thus mimicking the case of image data, transparently to the user.

Implementation of neural network for color properties of polycarbonates

NASA Astrophysics Data System (ADS)

Saeed, U.; Ahmad, S.; Alsadi, J.; Ross, D.; Rizvi, G.

2014-05-01

In present paper, the applicability of artificial neural networks (ANN) is investigated for color properties of plastics. The neural networks toolbox of Matlab 6.5 is used to develop and test the ANN model on a personal computer. An optimal design is completed for 10, 12, 14,16,18 & 20 hidden neurons on single hidden layer with five different algorithms: batch gradient descent (GD), batch variable learning rate (GDX), resilient back-propagation (RP), scaled conjugate gradient (SCG), levenberg-marquardt (LM) in the feed forward back-propagation neural network model. The training data for ANN is obtained from experimental measurements. There were twenty two inputs including resins, additives & pigments while three tristimulus color values L*, a* and b* were used as output layer. Statistical analysis in terms of Root-Mean-Squared (RMS), absolute fraction of variance (R squared), as well as mean square error is used to investigate the performance of ANN. LM algorithm with fourteen neurons on hidden layer in Feed Forward Back-Propagation of ANN model has shown best result in the present study. The degree of accuracy of the ANN model in reduction of errors is proven acceptable in all statistical analysis and shown in results. However, it was concluded that ANN provides a feasible method in error reduction in specific color tristimulus values.

Application of neural networks to prediction of advanced composite structures mechanical response and behavior

NASA Technical Reports Server (NTRS)

Cios, K. J.; Vary, A.; Berke, L.; Kautz, H. E.

1992-01-01

Two types of neural networks were used to evaluate acousto-ultrasonic (AU) data for material characterization and mechanical reponse prediction. The neural networks included a simple feedforward network (backpropagation) and a radial basis functions network. Comparisons of results in terms of accuracy and training time are given. Acousto-ultrasonic (AU) measurements were performed on a series of tensile specimens composed of eight laminated layers of continuous, SiC fiber reinforced Ti-15-3 matrix. The frequency spectrum was dominated by frequencies of longitudinal wave resonance through the thickness of the specimen at the sending transducer. The magnitude of the frequency spectrum of the AU signal was used for calculating a stress-wave factor based on integrating the spectral distribution function and used for comparison with neural networks results.

Single image super-resolution based on convolutional neural networks

NASA Astrophysics Data System (ADS)

Zou, Lamei; Luo, Ming; Yang, Weidong; Li, Peng; Jin, Liujia

2018-03-01

We present a deep learning method for single image super-resolution (SISR). The proposed approach learns end-to-end mapping between low-resolution (LR) images and high-resolution (HR) images. The mapping is represented as a deep convolutional neural network which inputs the LR image and outputs the HR image. Our network uses 5 convolution layers, which kernels size include 5×5, 3×3 and 1×1. In our proposed network, we use residual-learning and combine different sizes of convolution kernels at the same layer. The experiment results show that our proposed method performs better than the existing methods in reconstructing quality index and human visual effects on benchmarked images.

The application of improved neural network in hydrocarbon reservoir prediction

NASA Astrophysics Data System (ADS)

Peng, Xiaobo

2013-03-01

This paper use BP neural network techniques to realize hydrocarbon reservoir predication easier and faster in tarim basin in oil wells. A grey - cascade neural network model is proposed and it is faster convergence speed and low error rate. The new method overcomes the shortcomings of traditional BP neural network convergence slow, easy to achieve extreme minimum value. This study had 220 sets of measured logging data to the sample data training mode. By changing the neuron number and types of the transfer function of hidden layers, the best work prediction model is analyzed. The conclusion is the model which can produce good prediction results in general, and can be used for hydrocarbon reservoir prediction.

Deep multi-scale convolutional neural network for hyperspectral image classification

NASA Astrophysics Data System (ADS)

Zhang, Feng-zhe; Yang, Xia

2018-04-01

In this paper, we proposed a multi-scale convolutional neural network for hyperspectral image classification task. Firstly, compared with conventional convolution, we utilize multi-scale convolutions, which possess larger respective fields, to extract spectral features of hyperspectral image. We design a deep neural network with a multi-scale convolution layer which contains 3 different convolution kernel sizes. Secondly, to avoid overfitting of deep neural network, dropout is utilized, which randomly sleeps neurons, contributing to improve the classification accuracy a bit. In addition, new skills like ReLU in deep learning is utilized in this paper. We conduct experiments on University of Pavia and Salinas datasets, and obtained better classification accuracy compared with other methods.

Lateral Information Processing by Spiking Neurons: A Theoretical Model of the Neural Correlate of Consciousness

PubMed Central

Ebner, Marc; Hameroff, Stuart

2011-01-01

Cognitive brain functions, for example, sensory perception, motor control and learning, are understood as computation by axonal-dendritic chemical synapses in networks of integrate-and-fire neurons. Cognitive brain functions may occur either consciously or nonconsciously (on “autopilot”). Conscious cognition is marked by gamma synchrony EEG, mediated largely by dendritic-dendritic gap junctions, sideways connections in input/integration layers. Gap-junction-connected neurons define a sub-network within a larger neural network. A theoretical model (the “conscious pilot”) suggests that as gap junctions open and close, a gamma-synchronized subnetwork, or zone moves through the brain as an executive agent, converting nonconscious “auto-pilot” cognition to consciousness, and enhancing computation by coherent processing and collective integration. In this study we implemented sideways “gap junctions” in a single-layer artificial neural network to perform figure/ground separation. The set of neurons connected through gap junctions form a reconfigurable resistive grid or sub-network zone. In the model, outgoing spikes are temporally integrated and spatially averaged using the fixed resistive grid set up by neurons of similar function which are connected through gap-junctions. This spatial average, essentially a feedback signal from the neuron's output, determines whether particular gap junctions between neurons will open or close. Neurons connected through open gap junctions synchronize their output spikes. We have tested our gap-junction-defined sub-network in a one-layer neural network on artificial retinal inputs using real-world images. Our system is able to perform figure/ground separation where the laterally connected sub-network of neurons represents a perceived object. Even though we only show results for visual stimuli, our approach should generalize to other modalities. The system demonstrates a moving sub-network zone of synchrony, within which the contents of perception are represented and contained. This mobile zone can be viewed as a model of the neural correlate of consciousness in the brain. PMID:22046178

Lateral information processing by spiking neurons: a theoretical model of the neural correlate of consciousness.

PubMed

Ebner, Marc; Hameroff, Stuart

2011-01-01

Cognitive brain functions, for example, sensory perception, motor control and learning, are understood as computation by axonal-dendritic chemical synapses in networks of integrate-and-fire neurons. Cognitive brain functions may occur either consciously or nonconsciously (on "autopilot"). Conscious cognition is marked by gamma synchrony EEG, mediated largely by dendritic-dendritic gap junctions, sideways connections in input/integration layers. Gap-junction-connected neurons define a sub-network within a larger neural network. A theoretical model (the "conscious pilot") suggests that as gap junctions open and close, a gamma-synchronized subnetwork, or zone moves through the brain as an executive agent, converting nonconscious "auto-pilot" cognition to consciousness, and enhancing computation by coherent processing and collective integration. In this study we implemented sideways "gap junctions" in a single-layer artificial neural network to perform figure/ground separation. The set of neurons connected through gap junctions form a reconfigurable resistive grid or sub-network zone. In the model, outgoing spikes are temporally integrated and spatially averaged using the fixed resistive grid set up by neurons of similar function which are connected through gap-junctions. This spatial average, essentially a feedback signal from the neuron's output, determines whether particular gap junctions between neurons will open or close. Neurons connected through open gap junctions synchronize their output spikes. We have tested our gap-junction-defined sub-network in a one-layer neural network on artificial retinal inputs using real-world images. Our system is able to perform figure/ground separation where the laterally connected sub-network of neurons represents a perceived object. Even though we only show results for visual stimuli, our approach should generalize to other modalities. The system demonstrates a moving sub-network zone of synchrony, within which the contents of perception are represented and contained. This mobile zone can be viewed as a model of the neural correlate of consciousness in the brain.

Semantic Segmentation of Convolutional Neural Network for Supervised Classification of Multispectral Remote Sensing

NASA Astrophysics Data System (ADS)

Xue, L.; Liu, C.; Wu, Y.; Li, H.

2018-04-01

Semantic segmentation is a fundamental research in remote sensing image processing. Because of the complex maritime environment, the classification of roads, vegetation, buildings and water from remote Sensing Imagery is a challenging task. Although the neural network has achieved excellent performance in semantic segmentation in the last years, there are a few of works using CNN for ground object segmentation and the results could be further improved. This paper used convolution neural network named U-Net, its structure has a contracting path and an expansive path to get high resolution output. In the network , We added BN layers, which is more conducive to the reverse pass. Moreover, after upsampling convolution , we add dropout layers to prevent overfitting. They are promoted to get more precise segmentation results. To verify this network architecture, we used a Kaggle dataset. Experimental results show that U-Net achieved good performance compared with other architectures, especially in high-resolution remote sensing imagery.

Classification of E-Nose Aroma Data of Four Fruit Types by ABC-Based Neural Network

PubMed Central

Adak, M. Fatih; Yumusak, Nejat

2016-01-01

Electronic nose technology is used in many areas, and frequently in the beverage industry for classification and quality-control purposes. In this study, four different aroma data (strawberry, lemon, cherry, and melon) were obtained using a MOSES II electronic nose for the purpose of fruit classification. To improve the performance of the classification, the training phase of the neural network with two hidden layers was optimized using artificial bee colony algorithm (ABC), which is known to be successful in exploration. Test data were given to two different neural networks, each of which were trained separately with backpropagation (BP) and ABC, and average test performances were measured as 60% for the artificial neural network trained with BP and 76.39% for the artificial neural network trained with ABC. Training and test phases were repeated 30 times to obtain these average performance measurements. This level of performance shows that the artificial neural network trained with ABC is successful in classifying aroma data. PMID:26927124

Using Neural Networks to Describe Tracer Correlations

NASA Technical Reports Server (NTRS)

Lary, D. J.; Mueller, M. D.; Mussa, H. Y.

2003-01-01

Neural networks are ideally suited to describe the spatial and temporal dependence of tracer-tracer correlations. The neural network performs well even in regions where the correlations are less compact and normally a family of correlation curves would be required. For example, the CH4-N2O correlation can be well described using a neural network trained with the latitude, pressure, time of year, and CH4 volume mixing ratio (v.m.r.). In this study a neural network using Quickprop learning and one hidden layer with eight nodes was able to reproduce the CH4-N2O correlation with a correlation co- efficient of 0.9995. Such an accurate representation of tracer-tracer correlations allows more use to be made of long-term datasets to constrain chemical models. Such as the dataset from the Halogen Occultation Experiment (HALOE) which has continuously observed CH4, (but not N2O) from 1991 till the present. The neural network Fortran code used is available for download.

Classification of E-Nose Aroma Data of Four Fruit Types by ABC-Based Neural Network.

PubMed

Adak, M Fatih; Yumusak, Nejat

2016-02-27

Electronic nose technology is used in many areas, and frequently in the beverage industry for classification and quality-control purposes. In this study, four different aroma data (strawberry, lemon, cherry, and melon) were obtained using a MOSES II electronic nose for the purpose of fruit classification. To improve the performance of the classification, the training phase of the neural network with two hidden layers was optimized using artificial bee colony algorithm (ABC), which is known to be successful in exploration. Test data were given to two different neural networks, each of which were trained separately with backpropagation (BP) and ABC, and average test performances were measured as 60% for the artificial neural network trained with BP and 76.39% for the artificial neural network trained with ABC. Training and test phases were repeated 30 times to obtain these average performance measurements. This level of performance shows that the artificial neural network trained with ABC is successful in classifying aroma data.

[Prognosis of the IVF ICSI/ET procedure efficiency with the use of artificial neural networks among patients of the Department of Reproduction and Gynecological Endocrinology].

PubMed

Milewski, Robert; Jamiołkowski, Jacek; Milewska Anna, Justyna; Domitrz, Jan; Szamatowicz, Jacek; Wołczyński, Sławomir

2009-12-01

Prognosis of pregnancy for patients treated with IVF ICSI/ET methods, using artificial neural networks. Retrospective study of 1007 cycles of infertility treatment of 899 patients of Department of Reproduction and Gynecological Endocrinology in Bialystok. The subjects were treated with IVF ICSI/ET method from August 2005 to September 2008. Classifying artificial neural network is described in the paper Architecture of the network is three-layered perceptron consisting of 45 neurons in the input layer 14 neurons in the hidden layer and a single output neuron. The source data for the network are 36 variables. 24 of them are nominal variables and the rest are quantitative variables. Among non-pregnancy cases only 59 prognosis of the network were incorrect. The results of treatment were correctly forecast in 68.5% of cases. The pregnancy was accurately confirmed in 49.1% of cases and lack of pregnancy in 86.5% of cases. Treatment of infertility with the use of in vitro fertilization methods continues to have too low efficiency per one treatment cycle. To improve this indicator it is necessary to find dependencies, which describe the model of IVF treatment. The application of advanced methods of bioinformatics allows to predict the result of the treatment more effectively With the help of artificial neural networks, we are able to forecast the failure of the treatment using IFV ICSI/ET procedure with almost 90% probability of certainty These possibilities can be used to predict negative cases.

Bringing Interpretability and Visualization with Artificial Neural Networks

ERIC Educational Resources Information Center

Gritsenko, Andrey

2017-01-01

Extreme Learning Machine (ELM) is a training algorithm for Single-Layer Feed-forward Neural Network (SLFN). The difference in theory of ELM from other training algorithms is in the existence of explicitly-given solution due to the immutability of initialed weights. In practice, ELMs achieve performance similar to that of other state-of-the-art…

Diagonal recurrent neural network based adaptive control of nonlinear dynamical systems using lyapunov stability criterion.

PubMed

Kumar, Rajesh; Srivastava, Smriti; Gupta, J R P

2017-03-01

In this paper adaptive control of nonlinear dynamical systems using diagonal recurrent neural network (DRNN) is proposed. The structure of DRNN is a modification of fully connected recurrent neural network (FCRNN). Presence of self-recurrent neurons in the hidden layer of DRNN gives it an ability to capture the dynamic behaviour of the nonlinear plant under consideration (to be controlled). To ensure stability, update rules are developed using lyapunov stability criterion. These rules are then used for adjusting the various parameters of DRNN. The responses of plants obtained with DRNN are compared with those obtained when multi-layer feed forward neural network (MLFFNN) is used as a controller. Also, in example 4, FCRNN is also investigated and compared with DRNN and MLFFNN. Robustness of the proposed control scheme is also tested against parameter variations and disturbance signals. Four simulation examples including one-link robotic manipulator and inverted pendulum are considered on which the proposed controller is applied. The results so obtained show the superiority of DRNN over MLFFNN as a controller. Copyright © 2017 ISA. Published by Elsevier Ltd. All rights reserved.

Artificial neural network with backpropagation learning to predict mean monthly total ozone in Arosa, Switzerland

NASA Astrophysics Data System (ADS)

Chattopadhyay, Surajit; Bandyopadhyay, Goutami

2007-01-01

Present study deals with the mean monthly total ozone time series over Arosa, Switzerland. The study period is 1932-1971. First of all, the total ozone time series has been identified as a complex system and then Artificial Neural Networks models in the form of Multilayer Perceptron with back propagation learning have been developed. The models are Single-hidden-layer and Two-hidden-layer Perceptrons with sigmoid activation function. After sequential learning with learning rate 0.9 the peak total ozone period (February-May) concentrations of mean monthly total ozone have been predicted by the two neural net models. After training and validation, both of the models are found skillful. But, Two-hidden-layer Perceptron is found to be more adroit in predicting the mean monthly total ozone concentrations over the aforesaid period.

Prediction of properties of wheat dough using intelligent deep belief networks

NASA Astrophysics Data System (ADS)

Guha, Paramita; Bhatnagar, Taru; Pal, Ishan; Kamboj, Uma; Mishra, Sunita

2017-11-01

In this paper, the rheological and chemical properties of wheat dough are predicted using deep belief networks. Wheat grains are stored at controlled environmental conditions. The internal parameters of grains viz., protein, fat, carbohydrates, moisture, ash are determined using standard chemical analysis and viscosity of the dough is measured using Rheometer. Here, fat, carbohydrates, moisture, ash and temperature are considered as inputs whereas protein and viscosity are chosen as outputs. The prediction algorithm is developed using deep neural network where each layer is trained greedily using restricted Boltzmann machine (RBM) networks. The overall network is finally fine-tuned using standard neural network technique. In most literature, it has been found that fine-tuning is done using back-propagation technique. In this paper, a new algorithm is proposed in which each layer is tuned using RBM and the final network is fine-tuned using deep neural network (DNN). It has been observed that with the proposed algorithm, errors between the actual and predicted outputs are less compared to the conventional algorithm. Hence, the given network can be considered as beneficial as it predicts the outputs more accurately. Numerical results along with discussions are presented.

A neural network approach for image reconstruction in electron magnetic resonance tomography.

PubMed

Durairaj, D Christopher; Krishna, Murali C; Murugesan, Ramachandran

2007-10-01

An object-oriented, artificial neural network (ANN) based, application system for reconstruction of two-dimensional spatial images in electron magnetic resonance (EMR) tomography is presented. The standard back propagation algorithm is utilized to train a three-layer sigmoidal feed-forward, supervised, ANN to perform the image reconstruction. The network learns the relationship between the 'ideal' images that are reconstructed using filtered back projection (FBP) technique and the corresponding projection data (sinograms). The input layer of the network is provided with a training set that contains projection data from various phantoms as well as in vivo objects, acquired from an EMR imager. Twenty five different network configurations are investigated to test the ability of the generalization of the network. The trained ANN then reconstructs two-dimensional temporal spatial images that present the distribution of free radicals in biological systems. Image reconstruction by the trained neural network shows better time complexity than the conventional iterative reconstruction algorithms such as multiplicative algebraic reconstruction technique (MART). The network is further explored for image reconstruction from 'noisy' EMR data and the results show better performance than the FBP method. The network is also tested for its ability to reconstruct from limited-angle EMR data set.

Modeling constitutive behavior of a 15Cr-15Ni-2.2Mo-Ti modified austenitic stainless steel under hot compression using artificial neural network

NASA Astrophysics Data System (ADS)

Mandal, Sumantra

2006-11-01

ABSTRACT In this paper, an artificial neural network (ANN) model has been suggested to predict the constitutive flow behavior of a 15Cr-15Ni-2.2Mo-Ti modified austenitic stainless steel under hot deformation. Hot compression tests in the temperature range 850°C- 1250°C and strain rate range 10-3-102 s-1 were carried out. These tests provided the required data for training the neural network and for subsequent testing. The inputs of the neural network are strain, log strain rate and temperature while flow stress is obtained as output. A three layer feed-forward network with ten neurons in a single hidden layer and back-propagation learning algorithm has been employed. A very good correlation between experimental and predicted result has been obtained. The effect of temperature and strain rate on flow behavior has been simulated employing the ANN model. The results have been found to be consistent with the metallurgical trend. Finally, a monte carlo analiysis has been carried out to find out the noise sensitivity of the developed model.

Conjugate-Gradient Neural Networks in Classification of Multisource and Very-High-Dimensional Remote Sensing Data

NASA Technical Reports Server (NTRS)

Benediktsson, J. A.; Swain, P. H.; Ersoy, O. K.

1993-01-01

Application of neural networks to classification of remote sensing data is discussed. Conventional two-layer backpropagation is found to give good results in classification of remote sensing data but is not efficient in training. A more efficient variant, based on conjugate-gradient optimization, is used for classification of multisource remote sensing and geographic data and very-high-dimensional data. The conjugate-gradient neural networks give excellent performance in classification of multisource data, but do not compare as well with statistical methods in classification of very-high-dimentional data.

Artificial Intelligence in Prediction of Secondary Protein Structure Using CB513 Database

PubMed Central

Avdagic, Zikrija; Purisevic, Elvir; Omanovic, Samir; Coralic, Zlatan

2009-01-01

In this paper we describe CB513 a non-redundant dataset, suitable for development of algorithms for prediction of secondary protein structure. A program was made in Borland Delphi for transforming data from our dataset to make it suitable for learning of neural network for prediction of secondary protein structure implemented in MATLAB Neural-Network Toolbox. Learning (training and testing) of neural network is researched with different sizes of windows, different number of neurons in the hidden layer and different number of training epochs, while using dataset CB513. PMID:21347158

Optical waveguides with memory effect using photochromic material for neural network

NASA Astrophysics Data System (ADS)

Tanimoto, Keisuke; Amemiya, Yoshiteru; Yokoyama, Shin

2018-04-01

An optical neural network using a waveguide with a memory effect, a photodiode, CMOS circuits and LEDs was proposed. To realize the neural network, optical waveguides with a memory effect were fabricated using a cladding layer containing the photochromic material “diarylethene”. The transmittance of green light was decreased by UV light irradiation and recovered by the passage of green light through the waveguide. It was confirmed that the transmittance versus total energy of the green light that passed through the waveguide well fit the universal exponential curve.

Vanishing points detection using combination of fast Hough transform and deep learning

NASA Astrophysics Data System (ADS)

Sheshkus, Alexander; Ingacheva, Anastasia; Nikolaev, Dmitry

2018-04-01

In this paper we propose a novel method for vanishing points detection based on convolutional neural network (CNN) approach and fast Hough transform algorithm. We show how to determine fast Hough transform neural network layer and how to use it in order to increase usability of the neural network approach to the vanishing point detection task. Our algorithm includes CNN with consequence of convolutional and fast Hough transform layers. We are building estimator for distribution of possible vanishing points in the image. This distribution can be used to find candidates of vanishing point. We provide experimental results from tests of suggested method using images collected from videos of road trips. Our approach shows stable result on test images with different projective distortions and noise. Described approach can be effectively implemented for mobile GPU and CPU.

Ensemble learning in fixed expansion layer networks for mitigating catastrophic forgetting.

PubMed

Coop, Robert; Mishtal, Aaron; Arel, Itamar

2013-10-01

Catastrophic forgetting is a well-studied attribute of most parameterized supervised learning systems. A variation of this phenomenon, in the context of feedforward neural networks, arises when nonstationary inputs lead to loss of previously learned mappings. The majority of the schemes proposed in the literature for mitigating catastrophic forgetting were not data driven and did not scale well. We introduce the fixed expansion layer (FEL) feedforward neural network, which embeds a sparsely encoding hidden layer to help mitigate forgetting of prior learned representations. In addition, we investigate a novel framework for training ensembles of FEL networks, based on exploiting an information-theoretic measure of diversity between FEL learners, to further control undesired plasticity. The proposed methodology is demonstrated on a basic classification task, clearly emphasizing its advantages over existing techniques. The architecture proposed can be enhanced to address a range of computational intelligence tasks, such as regression problems and system control.

Neural network evaluation of reflectometry density profiles for control purposes

NASA Astrophysics Data System (ADS)

Santos, J.; Nunes, F.; Manso, M.; Nunes, I.

1999-01-01

Broadband reflectometry is a diagnostic that is able to measure the density profile with high spatial and temporal resolutions, therefore it can be used to improve the performance of advanced tokamak operation modes and to supplement or correct the magnetics for plasma position control. To perform these tasks real-time processing is needed. Here we present a method that uses a neural network to make a fast evaluation of radial positions for selected density layers. Typical ASDEX Upgrade density profiles were used to generate the simulated network training and test sets. It is shown that the method has the potential to meet the tight timing requirements of control applications with the required accuracy. The network is also able to provide an accurate estimation of the position of density layers below the first density layer which is probed by an O-mode reflectometer, provided that it is trained with a realistic density profile model.

Goal-seeking neural net for recall and recognition

NASA Astrophysics Data System (ADS)

Omidvar, Omid M.

1990-07-01

Neural networks have been used to mimic cognitive processes which take place in animal brains. The learning capability inherent in neural networks makes them suitable candidates for adaptive tasks such as recall and recognition. The synaptic reinforcements create a proper condition for adaptation, which results in memorization, formation of perception, and higher order information processing activities. In this research a model of a goal seeking neural network is studied and the operation of the network with regard to recall and recognition is analyzed. In these analyses recall is defined as retrieval of stored information where little or no matching is involved. On the other hand recognition is recall with matching; therefore it involves memorizing a piece of information with complete presentation. This research takes the generalized view of reinforcement in which all the signals are potential reinforcers. The neuronal response is considered to be the source of the reinforcement. This local approach to adaptation leads to the goal seeking nature of the neurons as network components. In the proposed model all the synaptic strengths are reinforced in parallel while the reinforcement among the layers is done in a distributed fashion and pipeline mode from the last layer inward. A model of complex neuron with varying threshold is developed to account for inhibitory and excitatory behavior of real neuron. A goal seeking model of a neural network is presented. This network is utilized to perform recall and recognition tasks. The performance of the model with regard to the assigned tasks is presented.

[Rapid Identification of Epicarpium Citri Grandis via Infrared Spectroscopy and Fluorescence Spectrum Imaging Technology Combined with Neural Network].

PubMed

Pan, Sha-sha; Huang, Fu-rong; Xiao, Chi; Xian, Rui-yi; Ma, Zhi-guo

2015-10-01

To explore rapid reliable methods for detection of Epicarpium citri grandis (ECG), the experiment using Fourier Transform Attenuated Total Reflection Infrared Spectroscopy (FTIR/ATR) and Fluorescence Spectrum Imaging Technology combined with Multilayer Perceptron (MLP) Neural Network pattern recognition, for the identification of ECG, and the two methods are compared. Infrared spectra and fluorescence spectral images of 118 samples, 81 ECG and 37 other kinds of ECG, are collected. According to the differences in tspectrum, the spectra data in the 550-1 800 cm(-1) wavenumber range and 400-720 nm wavelength are regarded as the study objects of discriminant analysis. Then principal component analysis (PCA) is applied to reduce the dimension of spectroscopic data of ECG and MLP Neural Network is used in combination to classify them. During the experiment were compared the effects of different methods of data preprocessing on the model: multiplicative scatter correction (MSC), standard normal variable correction (SNV), first-order derivative(FD), second-order derivative(SD) and Savitzky-Golay (SG). The results showed that: after the infrared spectra data via the Savitzky-Golay (SG) pretreatment through the MLP Neural Network with the hidden layer function as sigmoid, we can get the best discrimination of ECG, the correct percent of training set and testing set are both 100%. Using fluorescence spectral imaging technology, corrected by the multiple scattering (MSC) results in the pretreatment is the most ideal. After data preprocessing, the three layers of the MLP Neural Network of the hidden layer function as sigmoid function can get 100% correct percent of training set and 96.7% correct percent of testing set. It was shown that the FTIR/ATR and fluorescent spectral imaging technology combined with MLP Neural Network can be used for the identification study of ECG and has the advantages of rapid, reliable effect.

Application of Artificial Neural Network to Predict the use of Runway at Juanda International Airport

NASA Astrophysics Data System (ADS)

Putra, J. C. P.; Safrilah

2017-06-01

Artificial neural network approaches are useful to solve many complicated problems. It solves a number of problems in various areas such as engineering, medicine, business, manufacturing, etc. This paper presents an application of artificial neural network to predict a runway capacity at Juanda International Airport. An artificial neural network model of backpropagation and multi-layer perceptron is adopted to this research to learning process of runway capacity at Juanda International Airport. The results indicate that the training data is successfully recognizing the certain pattern of runway use at Juanda International Airport. Whereas, testing data indicate vice versa. Finally, it can be concluded that the approach of uniformity data and network architecture is the critical part to determine the accuracy of prediction results.

Deep neural network for traffic sign recognition systems: An analysis of spatial transformers and stochastic optimisation methods.

PubMed

Arcos-García, Álvaro; Álvarez-García, Juan A; Soria-Morillo, Luis M

2018-03-01

This paper presents a Deep Learning approach for traffic sign recognition systems. Several classification experiments are conducted over publicly available traffic sign datasets from Germany and Belgium using a Deep Neural Network which comprises Convolutional layers and Spatial Transformer Networks. Such trials are built to measure the impact of diverse factors with the end goal of designing a Convolutional Neural Network that can improve the state-of-the-art of traffic sign classification task. First, different adaptive and non-adaptive stochastic gradient descent optimisation algorithms such as SGD, SGD-Nesterov, RMSprop and Adam are evaluated. Subsequently, multiple combinations of Spatial Transformer Networks placed at distinct positions within the main neural network are analysed. The recognition rate of the proposed Convolutional Neural Network reports an accuracy of 99.71% in the German Traffic Sign Recognition Benchmark, outperforming previous state-of-the-art methods and also being more efficient in terms of memory requirements. Copyright © 2018 Elsevier Ltd. All rights reserved.

Face recognition: a convolutional neural-network approach.

PubMed

Lawrence, S; Giles, C L; Tsoi, A C; Back, A D

1997-01-01

We present a hybrid neural-network for human face recognition which compares favourably with other methods. The system combines local image sampling, a self-organizing map (SOM) neural network, and a convolutional neural network. The SOM provides a quantization of the image samples into a topological space where inputs that are nearby in the original space are also nearby in the output space, thereby providing dimensionality reduction and invariance to minor changes in the image sample, and the convolutional neural network provides partial invariance to translation, rotation, scale, and deformation. The convolutional network extracts successively larger features in a hierarchical set of layers. We present results using the Karhunen-Loeve transform in place of the SOM, and a multilayer perceptron (MLP) in place of the convolutional network for comparison. We use a database of 400 images of 40 individuals which contains quite a high degree of variability in expression, pose, and facial details. We analyze the computational complexity and discuss how new classes could be added to the trained recognizer.

Prediction of friction factor of pure water flowing inside vertical smooth and microfin tubes by using artificial neural networks

NASA Astrophysics Data System (ADS)

Çebi, A.; Akdoğan, E.; Celen, A.; Dalkilic, A. S.

2017-02-01

An artificial neural network (ANN) model of friction factor in smooth and microfin tubes under heating, cooling and isothermal conditions was developed in this study. Data used in ANN was taken from a vertically positioned heat exchanger experimental setup. Multi-layered feed-forward neural network with backpropagation algorithm, radial basis function networks and hybrid PSO-neural network algorithm were applied to the database. Inputs were the ratio of cross sectional flow area to hydraulic diameter, experimental condition number depending on isothermal, heating, or cooling conditions and mass flow rate while the friction factor was the output of the constructed system. It was observed that such neural network based system could effectively predict the friction factor values of the flows regardless of their tube types. A dependency analysis to determine the strongest parameter that affected the network and database was also performed and tube geometry was found to be the strongest parameter of all as a result of analysis.

Forecasting SPEI and SPI Drought Indices Using the Integrated Artificial Neural Networks

PubMed Central

Maca, Petr; Pech, Pavel

2016-01-01

The presented paper compares forecast of drought indices based on two different models of artificial neural networks. The first model is based on feedforward multilayer perceptron, sANN, and the second one is the integrated neural network model, hANN. The analyzed drought indices are the standardized precipitation index (SPI) and the standardized precipitation evaporation index (SPEI) and were derived for the period of 1948–2002 on two US catchments. The meteorological and hydrological data were obtained from MOPEX experiment. The training of both neural network models was made by the adaptive version of differential evolution, JADE. The comparison of models was based on six model performance measures. The results of drought indices forecast, explained by the values of four model performance indices, show that the integrated neural network model was superior to the feedforward multilayer perceptron with one hidden layer of neurons. PMID:26880875

Forecasting SPEI and SPI Drought Indices Using the Integrated Artificial Neural Networks.

PubMed

Maca, Petr; Pech, Pavel

2016-01-01

The presented paper compares forecast of drought indices based on two different models of artificial neural networks. The first model is based on feedforward multilayer perceptron, sANN, and the second one is the integrated neural network model, hANN. The analyzed drought indices are the standardized precipitation index (SPI) and the standardized precipitation evaporation index (SPEI) and were derived for the period of 1948-2002 on two US catchments. The meteorological and hydrological data were obtained from MOPEX experiment. The training of both neural network models was made by the adaptive version of differential evolution, JADE. The comparison of models was based on six model performance measures. The results of drought indices forecast, explained by the values of four model performance indices, show that the integrated neural network model was superior to the feedforward multilayer perceptron with one hidden layer of neurons.

A neural network for recognizing movement patterns during repetitive self-paced movements of the fingers in opposition to the thumb.

PubMed

Van Vaerenbergh, J; Vranken, R; Briers, L; Briers, H

2001-11-01

A data glove is a typical input device to control a virtual environment. At the same time it measures movements of wrist and fingers. The purposes of this investigation were to assess the ability of BrainMaker, a neural network, to recognize movement patterns during an opposition task that consisted of repetitive self-paced movements of the fingers in opposition to the thumb. The neural network contained 56 inputs, 3 hidden layers of 20 neurons, and one output. The 5th glove '95 (5DT), a commercial glove especially designed for virtual reality games, was used for finger motion capture. The training of the neural network was successful for recognizing the thumb, the index finger and the ring finger movements during the repetitive self-paced movements and neural network performed well during testing.

Application of Neural Networks for classification of Patau, Edwards, Down, Turner and Klinefelter Syndrome based on first trimester maternal serum screening data, ultrasonographic findings and patient demographics.

PubMed

Catic, Aida; Gurbeta, Lejla; Kurtovic-Kozaric, Amina; Mehmedbasic, Senad; Badnjevic, Almir

2018-02-13

The usage of Artificial Neural Networks (ANNs) for genome-enabled classifications and establishing genome-phenotype correlations have been investigated more extensively over the past few years. The reason for this is that ANNs are good approximates of complex functions, so classification can be performed without the need for explicitly defined input-output model. This engineering tool can be applied for optimization of existing methods for disease/syndrome classification. Cytogenetic and molecular analyses are the most frequent tests used in prenatal diagnostic for the early detection of Turner, Klinefelter, Patau, Edwards and Down syndrome. These procedures can be lengthy, repetitive; and often employ invasive techniques so a robust automated method for classifying and reporting prenatal diagnostics would greatly help the clinicians with their routine work. The database consisted of data collected from 2500 pregnant woman that came to the Institute of Gynecology, Infertility and Perinatology "Mehmedbasic" for routine antenatal care between January 2000 and December 2016. During first trimester all women were subject to screening test where values of maternal serum pregnancy-associated plasma protein A (PAPP-A) and free beta human chorionic gonadotropin (β-hCG) were measured. Also, fetal nuchal translucency thickness and the presence or absence of the nasal bone was observed using ultrasound. The architectures of linear feedforward and feedback neural networks were investigated for various training data distributions and number of neurons in hidden layer. Feedback neural network architecture out performed feedforward neural network architecture in predictive ability for all five aneuploidy prenatal syndrome classes. Feedforward neural network with 15 neurons in hidden layer achieved classification sensitivity of 92.00%. Classification sensitivity of feedback (Elman's) neural network was 99.00%. Average accuracy of feedforward neural network was 89.6% and for feedback was 98.8%. The results presented in this paper prove that an expert diagnostic system based on neural networks can be efficiently used for classification of five aneuploidy syndromes, covered with this study, based on first trimester maternal serum screening data, ultrasonographic findings and patient demographics. Developed Expert System proved to be simple, robust, and powerful in properly classifying prenatal aneuploidy syndromes.

The Use of Neural Network Technology to Model Swimming Performance

PubMed Central

Silva, António José; Costa, Aldo Manuel; Oliveira, Paulo Moura; Reis, Victor Machado; Saavedra, José; Perl, Jurgen; Rouboa, Abel; Marinho, Daniel Almeida

2007-01-01

The aims of the present study were: to identify the factors which are able to explain the performance in the 200 meters individual medley and 400 meters front crawl events in young swimmers, to model the performance in those events using non-linear mathematic methods through artificial neural networks (multi-layer perceptrons) and to assess the neural network models precision to predict the performance. A sample of 138 young swimmers (65 males and 73 females) of national level was submitted to a test battery comprising four different domains: kinanthropometric evaluation, dry land functional evaluation (strength and flexibility), swimming functional evaluation (hydrodynamics, hydrostatic and bioenergetics characteristics) and swimming technique evaluation. To establish a profile of the young swimmer non-linear combinations between preponderant variables for each gender and swim performance in the 200 meters medley and 400 meters font crawl events were developed. For this purpose a feed forward neural network was used (Multilayer Perceptron) with three neurons in a single hidden layer. The prognosis precision of the model (error lower than 0.8% between true and estimated performances) is supported by recent evidence. Therefore, we consider that the neural network tool can be a good approach in the resolution of complex problems such as performance modeling and the talent identification in swimming and, possibly, in a wide variety of sports. Key pointsThe non-linear analysis resulting from the use of feed forward neural network allowed us the development of four performance models.The mean difference between the true and estimated results performed by each one of the four neural network models constructed was low.The neural network tool can be a good approach in the resolution of the performance modeling as an alternative to the standard statistical models that presume well-defined distributions and independence among all inputs.The use of neural networks for sports sciences application allowed us to create very realistic models for swimming performance prediction based on previous selected criterions that were related with the dependent variable (performance). PMID:24149233

Neural network pattern recognition of thermal-signature spectra for chemical defense

NASA Astrophysics Data System (ADS)

Carrieri, Arthur H.; Lim, Pascal I.

1995-05-01

We treat infrared patterns of absorption or emission by nerve and blister agent compounds (and simulants of this chemical group) as features for the training of neural networks to detect the compounds' liquid layers on the ground or their vapor plumes during evaporation by external heating. Training of a four-layer network architecture is composed of a backward-error-propagation algorithm and a gradient-descent paradigm. We conduct testing by feed-forwarding preprocessed spectra through the network in a scaled format consistent with the structure of the training-data-set representation. The best-performance weight matrix (spectral filter) evolved from final network training and testing with software simulation trials is electronically transferred to a set of eight artificial intelligence integrated circuits (ICs') in specific modular form (splitting of weight matrices). This form makes full use of all input-output IC nodes. This neural network computer serves an important real-time detection function when it is integrated into pre-and postprocessing data-handling units of a tactical prototype thermoluminescence sensor now under development at the Edgewood Research, Development, and Engineering Center.

High Performance Implementation of 3D Convolutional Neural Networks on a GPU.

PubMed

Lan, Qiang; Wang, Zelong; Wen, Mei; Zhang, Chunyuan; Wang, Yijie

2017-01-01

Convolutional neural networks have proven to be highly successful in applications such as image classification, object tracking, and many other tasks based on 2D inputs. Recently, researchers have started to apply convolutional neural networks to video classification, which constitutes a 3D input and requires far larger amounts of memory and much more computation. FFT based methods can reduce the amount of computation, but this generally comes at the cost of an increased memory requirement. On the other hand, the Winograd Minimal Filtering Algorithm (WMFA) can reduce the number of operations required and thus can speed up the computation, without increasing the required memory. This strategy was shown to be successful for 2D neural networks. We implement the algorithm for 3D convolutional neural networks and apply it to a popular 3D convolutional neural network which is used to classify videos and compare it to cuDNN. For our highly optimized implementation of the algorithm, we observe a twofold speedup for most of the 3D convolution layers of our test network compared to the cuDNN version.

High Performance Implementation of 3D Convolutional Neural Networks on a GPU

PubMed Central

Wang, Zelong; Wen, Mei; Zhang, Chunyuan; Wang, Yijie

2017-01-01

Convolutional neural networks have proven to be highly successful in applications such as image classification, object tracking, and many other tasks based on 2D inputs. Recently, researchers have started to apply convolutional neural networks to video classification, which constitutes a 3D input and requires far larger amounts of memory and much more computation. FFT based methods can reduce the amount of computation, but this generally comes at the cost of an increased memory requirement. On the other hand, the Winograd Minimal Filtering Algorithm (WMFA) can reduce the number of operations required and thus can speed up the computation, without increasing the required memory. This strategy was shown to be successful for 2D neural networks. We implement the algorithm for 3D convolutional neural networks and apply it to a popular 3D convolutional neural network which is used to classify videos and compare it to cuDNN. For our highly optimized implementation of the algorithm, we observe a twofold speedup for most of the 3D convolution layers of our test network compared to the cuDNN version. PMID:29250109

Singularities of Three-Layered Complex-Valued Neural Networks With Split Activation Function.

PubMed

Kobayashi, Masaki

2018-05-01

There are three important concepts related to learning processes in neural networks: reducibility, nonminimality, and singularity. Although the definitions of these three concepts differ, they are equivalent in real-valued neural networks. This is also true of complex-valued neural networks (CVNNs) with hidden neurons not employing biases. The situation of CVNNs with hidden neurons employing biases, however, is very complicated. Exceptional reducibility was found, and it was shown that reducibility and nonminimality are not the same. Irreducibility consists of minimality and exceptional reducibility. The relationship between minimality and singularity has not yet been established. In this paper, we describe our surprising finding that minimality and singularity are independent. We also provide several examples based on exceptional reducibility.

Neural network controller development for a magnetically suspended flywheel energy storage system

NASA Technical Reports Server (NTRS)

Fittro, Roger L.; Pang, Da-Chen; Anand, Davinder K.

1994-01-01

A neural network controller has been developed to accommodate disturbances and nonlinearities and improve the robustness of a magnetically suspended flywheel energy storage system. The controller is trained using the back propagation-through-time technique incorporated with a time-averaging scheme. The resulting nonlinear neural network controller improves system performance by adapting flywheel stiffness and damping based on operating speed. In addition, a hybrid multi-layered neural network controller is developed off-line which is capable of improving system performance even further. All of the research presented in this paper was implemented via a magnetic bearing computer simulation. However, careful attention was paid to developing a practical methodology which will make future application to the actual bearing system fairly straightforward.

Constructing general partial differential equations using polynomial and neural networks.

PubMed

Zjavka, Ladislav; Pedrycz, Witold

2016-01-01

Sum fraction terms can approximate multi-variable functions on the basis of discrete observations, replacing a partial differential equation definition with polynomial elementary data relation descriptions. Artificial neural networks commonly transform the weighted sum of inputs to describe overall similarity relationships of trained and new testing input patterns. Differential polynomial neural networks form a new class of neural networks, which construct and solve an unknown general partial differential equation of a function of interest with selected substitution relative terms using non-linear multi-variable composite polynomials. The layers of the network generate simple and composite relative substitution terms whose convergent series combinations can describe partial dependent derivative changes of the input variables. This regression is based on trained generalized partial derivative data relations, decomposed into a multi-layer polynomial network structure. The sigmoidal function, commonly used as a nonlinear activation of artificial neurons, may transform some polynomial items together with the parameters with the aim to improve the polynomial derivative term series ability to approximate complicated periodic functions, as simple low order polynomials are not able to fully make up for the complete cycles. The similarity analysis facilitates substitutions for differential equations or can form dimensional units from data samples to describe real-world problems. Copyright © 2015 Elsevier Ltd. All rights reserved.

Path optimisation of a mobile robot using an artificial neural network controller

NASA Astrophysics Data System (ADS)

Singh, M. K.; Parhi, D. R.

2011-01-01

This article proposed a novel approach for design of an intelligent controller for an autonomous mobile robot using a multilayer feed forward neural network, which enables the robot to navigate in a real world dynamic environment. The inputs to the proposed neural controller consist of left, right and front obstacle distance with respect to its position and target angle. The output of the neural network is steering angle. A four layer neural network has been designed to solve the path and time optimisation problem of mobile robots, which deals with the cognitive tasks such as learning, adaptation, generalisation and optimisation. A back propagation algorithm is used to train the network. This article also analyses the kinematic design of mobile robots for dynamic movements. The simulation results are compared with experimental results, which are satisfactory and show very good agreement. The training of the neural nets and the control performance analysis has been done in a real experimental setup.

Application of Artificial Neural Network to Optical Fluid Analyzer

NASA Astrophysics Data System (ADS)

Kimura, Makoto; Nishida, Katsuhiko

1994-04-01

A three-layer artificial neural network has been applied to the presentation of optical fluid analyzer (OFA) raw data, and the accuracy of oil fraction determination has been significantly improved compared to previous approaches. To apply the artificial neural network approach to solving a problem, the first step is training to determine the appropriate weight set for calculating the target values. This involves using a series of data sets (each comprising a set of input values and an associated set of output values that the artificial neural network is required to determine) to tune artificial neural network weighting parameters so that the output of the neural network to the given set of input values is as close as possible to the required output. The physical model used to generate the series of learning data sets was the effective flow stream model, developed for OFA data presentation. The effectiveness of the training was verified by reprocessing the same input data as were used to determine the weighting parameters and then by comparing the results of the artificial neural network to the expected output values. The standard deviation of the expected and obtained values was approximately 10% (two sigma).

Multi-Layered Feedforward Neural Networks for Image Segmentation

DTIC Science & Technology

1991-12-01

the Gram-Schmidt Network ...................... 80 xi Preface WILLIAM SHAKESPEARE 1564-1616 Is this a dagger which I see before me, The handle toward...any input-output mapping with a single hidden layer of non-linear nodes, the result may be like proving that a monkey could write Hamlet . Certainly it

Planning Training Loads for the 400 M Hurdles in Three-Month Mesocycles using Artificial Neural Networks.

PubMed

Przednowek, Krzysztof; Iskra, Janusz; Wiktorowicz, Krzysztof; Krzeszowski, Tomasz; Maszczyk, Adam

2017-12-01

This paper presents a novel approach to planning training loads in hurdling using artificial neural networks. The neural models performed the task of generating loads for athletes' training for the 400 meters hurdles. All the models were calculated based on the training data of 21 Polish National Team hurdlers, aged 22.25 ± 1.96, competing between 1989 and 2012. The analysis included 144 training plans that represented different stages in the annual training cycle. The main contribution of this paper is to develop neural models for planning training loads for the entire career of a typical hurdler. In the models, 29 variables were used, where four characterized the runner and 25 described the training process. Two artificial neural networks were used: a multi-layer perceptron and a network with radial basis functions. To assess the quality of the models, the leave-one-out cross-validation method was used in which the Normalized Root Mean Squared Error was calculated. The analysis shows that the method generating the smallest error was the radial basis function network with nine neurons in the hidden layer. Most of the calculated training loads demonstrated a non-linear relationship across the entire competitive period. The resulting model can be used as a tool to assist a coach in planning training loads during a selected training period.

Planning Training Loads for the 400 M Hurdles in Three-Month Mesocycles using Artificial Neural Networks

PubMed Central

Iskra, Janusz; Wiktorowicz, Krzysztof; Krzeszowski, Tomasz; Maszczyk, Adam

2017-01-01

Abstract This paper presents a novel approach to planning training loads in hurdling using artificial neural networks. The neural models performed the task of generating loads for athletes’ training for the 400 meters hurdles. All the models were calculated based on the training data of 21 Polish National Team hurdlers, aged 22.25 ± 1.96, competing between 1989 and 2012. The analysis included 144 training plans that represented different stages in the annual training cycle. The main contribution of this paper is to develop neural models for planning training loads for the entire career of a typical hurdler. In the models, 29 variables were used, where four characterized the runner and 25 described the training process. Two artificial neural networks were used: a multi-layer perceptron and a network with radial basis functions. To assess the quality of the models, the leave-one-out cross-validation method was used in which the Normalized Root Mean Squared Error was calculated. The analysis shows that the method generating the smallest error was the radial basis function network with nine neurons in the hidden layer. Most of the calculated training loads demonstrated a non-linear relationship across the entire competitive period. The resulting model can be used as a tool to assist a coach in planning training loads during a selected training period. PMID:29339998

A Squeezed Artificial Neural Network for the Symbolic Network Reliability Functions of Binary-State Networks.

PubMed

Yeh, Wei-Chang

Network reliability is an important index to the provision of useful information for decision support in the modern world. There is always a need to calculate symbolic network reliability functions (SNRFs) due to dynamic and rapid changes in network parameters. In this brief, the proposed squeezed artificial neural network (SqANN) approach uses the Monte Carlo simulation to estimate the corresponding reliability of a given designed matrix from the Box-Behnken design, and then the Taguchi method is implemented to find the appropriate number of neurons and activation functions of the hidden layer and the output layer in ANN to evaluate SNRFs. According to the experimental results of the benchmark networks, the comparison appears to support the superiority of the proposed SqANN method over the traditional ANN-based approach with at least 16.6% improvement in the median absolute deviation in the cost of extra 2 s on average for all experiments.Network reliability is an important index to the provision of useful information for decision support in the modern world. There is always a need to calculate symbolic network reliability functions (SNRFs) due to dynamic and rapid changes in network parameters. In this brief, the proposed squeezed artificial neural network (SqANN) approach uses the Monte Carlo simulation to estimate the corresponding reliability of a given designed matrix from the Box-Behnken design, and then the Taguchi method is implemented to find the appropriate number of neurons and activation functions of the hidden layer and the output layer in ANN to evaluate SNRFs. According to the experimental results of the benchmark networks, the comparison appears to support the superiority of the proposed SqANN method over the traditional ANN-based approach with at least 16.6% improvement in the median absolute deviation in the cost of extra 2 s on average for all experiments.

Object recognition using deep convolutional neural networks with complete transfer and partial frozen layers

NASA Astrophysics Data System (ADS)

Kruithof, Maarten C.; Bouma, Henri; Fischer, Noëlle M.; Schutte, Klamer

2016-10-01

Object recognition is important to understand the content of video and allow flexible querying in a large number of cameras, especially for security applications. Recent benchmarks show that deep convolutional neural networks are excellent approaches for object recognition. This paper describes an approach of domain transfer, where features learned from a large annotated dataset are transferred to a target domain where less annotated examples are available as is typical for the security and defense domain. Many of these networks trained on natural images appear to learn features similar to Gabor filters and color blobs in the first layer. These first-layer features appear to be generic for many datasets and tasks while the last layer is specific. In this paper, we study the effect of copying all layers and fine-tuning a variable number. We performed an experiment with a Caffe-based network on 1000 ImageNet classes that are randomly divided in two equal subgroups for the transfer from one to the other. We copy all layers and vary the number of layers that is fine-tuned and the size of the target dataset. We performed additional experiments with the Keras platform on CIFAR-10 dataset to validate general applicability. We show with both platforms and both datasets that the accuracy on the target dataset improves when more target data is used. When the target dataset is large, it is beneficial to freeze only a few layers. For a large target dataset, the network without transfer learning performs better than the transfer network, especially if many layers are frozen. When the target dataset is small, it is beneficial to transfer (and freeze) many layers. For a small target dataset, the transfer network boosts generalization and it performs much better than the network without transfer learning. Learning time can be reduced by freezing many layers in a network.

Fault detection and classification in electrical power transmission system using artificial neural network.

PubMed

Jamil, Majid; Sharma, Sanjeev Kumar; Singh, Rajveer

2015-01-01

This paper focuses on the detection and classification of the faults on electrical power transmission line using artificial neural networks. The three phase currents and voltages of one end are taken as inputs in the proposed scheme. The feed forward neural network along with back propagation algorithm has been employed for detection and classification of the fault for analysis of each of the three phases involved in the process. A detailed analysis with varying number of hidden layers has been performed to validate the choice of the neural network. The simulation results concluded that the present method based on the neural network is efficient in detecting and classifying the faults on transmission lines with satisfactory performances. The different faults are simulated with different parameters to check the versatility of the method. The proposed method can be extended to the Distribution network of the Power System. The various simulations and analysis of signals is done in the MATLAB(®) environment.

A Deep Neural Network Model for Rainfall Estimation UsingPolarimetric WSR-88DP Radar Observations

NASA Astrophysics Data System (ADS)

Tan, H.; Chandra, C. V.; Chen, H.

2016-12-01

Rainfall estimation based on radar measurements has been an important topic for a few decades. Generally, radar rainfall estimation is conducted through parametric algorisms such as reflectivity-rainfall relation (i.e., Z-R relation). On the other hand, neural networks are developed for ground rainfall estimation based on radar measurements. This nonparametric method, which takes into account of both radar observations and rainfall measurements from ground rain gauges, has been demonstrated successfully for rainfall rate estimation. However, the neural network-based rainfall estimation is limited in practice due to the model complexity and structure, data quality, as well as different rainfall microphysics. Recently, the deep learning approach has been introduced in pattern recognition and machine learning areas. Compared to traditional neural networks, the deep learning based methodologies have larger number of hidden layers and more complex structure for data representation. Through a hierarchical learning process, the high level structured information and knowledge can be extracted automatically from low level features of the data. In this paper, we introduce a novel deep neural network model for rainfall estimation based on ground polarimetric radar measurements .The model is designed to capture the complex abstractions of radar measurements at different levels using multiple layers feature identification and extraction. The abstractions at different levels can be used independently or fused with other data resource such as satellite-based rainfall products and/or topographic data to represent the rain characteristics at certain location. In particular, the WSR-88DP radar and rain gauge data collected in Dallas - Fort Worth Metroplex and Florida are used extensively to train the model, and for demonstration purposes. Quantitative evaluation of the deep neural network based rainfall products will also be presented, which is based on an independent rain gauge network.

Neural network-based landmark detection for mobile robot

NASA Astrophysics Data System (ADS)

Sekiguchi, Minoru; Okada, Hiroyuki; Watanabe, Nobuo

1996-03-01

The mobile robot can essentially have only the relative position data for the real world. However, there are many cases that the robot has to know where it is located. In those cases, the useful method is to detect landmarks in the real world and adjust its position using detected landmarks. In this point of view, it is essential to develop a mobile robot that can accomplish the path plan successfully using natural or artificial landmarks. However, artificial landmarks are often difficult to construct and natural landmarks are very complicated to detect. In this paper, the method of acquiring landmarks by using the sensor data from the mobile robot necessary for planning the path is described. The landmark we discuss here is the natural one and is composed of the compression of sensor data from the robot. The sensor data is compressed and memorized by using five layered neural network that is called a sand glass model. The input and output data that neural network should learn is the sensor data of the robot that are exactly the same. Using the intermediate output data of the network, a compressed data is obtained, which expresses a landmark data. If the sensor data is ambiguous or enormous, it is easy to detect the landmark because the data is compressed and classified by the neural network. Using the backward three layers, the compressed landmark data is expanded to original data at some level. The studied neural network categorizes the detected sensor data to the known landmark.

Prediction of Sea Surface Temperature Using Long Short-Term Memory

NASA Astrophysics Data System (ADS)

Zhang, Qin; Wang, Hui; Dong, Junyu; Zhong, Guoqiang; Sun, Xin

2017-10-01

This letter adopts long short-term memory(LSTM) to predict sea surface temperature(SST), which is the first attempt, to our knowledge, to use recurrent neural network to solve the problem of SST prediction, and to make one week and one month daily prediction. We formulate the SST prediction problem as a time series regression problem. LSTM is a special kind of recurrent neural network, which introduces gate mechanism into vanilla RNN to prevent the vanished or exploding gradient problem. It has strong ability to model the temporal relationship of time series data and can handle the long-term dependency problem well. The proposed network architecture is composed of two kinds of layers: LSTM layer and full-connected dense layer. LSTM layer is utilized to model the time series relationship. Full-connected layer is utilized to map the output of LSTM layer to a final prediction. We explore the optimal setting of this architecture by experiments and report the accuracy of coastal seas of China to confirm the effectiveness of the proposed method. In addition, we also show its online updated characteristics.

Improved automatic adjustment of density and contrast in FCR system using neural network

NASA Astrophysics Data System (ADS)

Takeo, Hideya; Nakajima, Nobuyoshi; Ishida, Masamitsu; Kato, Hisatoyo

1994-05-01

FCR system has an automatic adjustment of image density and contrast by analyzing the histogram of image data in the radiation field. Advanced image recognition methods proposed in this paper can improve the automatic adjustment performance, in which neural network technology is used. There are two methods. Both methods are basically used 3-layer neural network with back propagation. The image data are directly input to the input-layer in one method and the histogram data is input in the other method. The former is effective to the imaging menu such as shoulder joint in which the position of interest region occupied on the histogram changes by difference of positioning and the latter is effective to the imaging menu such as chest-pediatrics in which the histogram shape changes by difference of positioning. We experimentally confirm the validity of these methods (about the automatic adjustment performance) as compared with the conventional histogram analysis methods.

Evaluation of thermal conductivity of MgO-MWCNTs/EG hybrid nanofluids based on experimental data by selecting optimal artificial neural networks

NASA Astrophysics Data System (ADS)

Vafaei, Masoud; Afrand, Masoud; Sina, Nima; Kalbasi, Rasool; Sourani, Forough; Teimouri, Hamid

2017-01-01

In this paper, the thermal conductivity ratio of MgO-MWCNTs/EG hybrid nanofluids has been predicted by an optimal artificial neural network at solid volume fractions of 0.05%, 0.1%, 0.15%, 0.2%, 0.4% and 0.6% in the temperature range of 25-50 °C. In this way, at the first, thirty six experimental data was presented to determine the thermal conductivity ratio of the hybrid nanofluid. Then, four optimal artificial neural networks with 6, 8, 10 and 12 neurons in hidden layer were designed to predict the thermal conductivity ratio of the nanofluid. The comparison between four optimal ANN results and experimental showed that the ANN with 12 neurons in hidden layer was the best model. Moreover, the results obtained from the best ANN indicated the maximum deviation margin of 0.8%.

Application of dynamic recurrent neural networks in nonlinear system identification

NASA Astrophysics Data System (ADS)

Du, Yun; Wu, Xueli; Sun, Huiqin; Zhang, Suying; Tian, Qiang

2006-11-01

An adaptive identification method of simple dynamic recurrent neural network (SRNN) for nonlinear dynamic systems is presented in this paper. This method based on the theory that by using the inner-states feed-back of dynamic network to describe the nonlinear kinetic characteristics of system can reflect the dynamic characteristics more directly, deduces the recursive prediction error (RPE) learning algorithm of SRNN, and improves the algorithm by studying topological structure on recursion layer without the weight values. The simulation results indicate that this kind of neural network can be used in real-time control, due to its less weight values, simpler learning algorithm, higher identification speed, and higher precision of model. It solves the problems of intricate in training algorithm and slow rate in convergence caused by the complicate topological structure in usual dynamic recurrent neural network.

An implementation of Elman neural network for polycystic ovary classification based on ultrasound images

NASA Astrophysics Data System (ADS)

Thufailah, I. F.; Adiwijaya; Wisesty, U. N.; Jondri

2018-03-01

Polycystic Ovary Syndrome (PCOS) is a reproduction problem that causes irregular menstruation period. Insulin and androgen hormone have big roles for this problem. This syndrome should be detected shortly, since it is able to cause a more serious disease, such as cardiovascular, diabetes, and obesity. The detection of this syndrome is done by analyzing ovary morphology and hormone test. However, the more economical way of test is by identifying the ovary morphology using ultrasonography. To classify whether one ovary is normal or it has polycystic ovary (PCO) follicle, the analysis will be done manually by a gynecologist. This paper will design a system to detect PCO using Gabor Wavelet method for feature extraction and Elman Neural Network is used to classify PCO and non-PCO. Elman Neural Network is chosen because it contains context layer to recall the previous condition. This paper compared the accuracy and process time of each dataset, then also did testing on elman’s parameters, such as layer delay, hidden layer, and training function. Based on tests done in this paper, the most accurate number is 78.1% with 32 features.

Object recognition with hierarchical discriminant saliency networks.

PubMed

Han, Sunhyoung; Vasconcelos, Nuno

2014-01-01

The benefits of integrating attention and object recognition are investigated. While attention is frequently modeled as a pre-processor for recognition, we investigate the hypothesis that attention is an intrinsic component of recognition and vice-versa. This hypothesis is tested with a recognition model, the hierarchical discriminant saliency network (HDSN), whose layers are top-down saliency detectors, tuned for a visual class according to the principles of discriminant saliency. As a model of neural computation, the HDSN has two possible implementations. In a biologically plausible implementation, all layers comply with the standard neurophysiological model of visual cortex, with sub-layers of simple and complex units that implement a combination of filtering, divisive normalization, pooling, and non-linearities. In a convolutional neural network implementation, all layers are convolutional and implement a combination of filtering, rectification, and pooling. The rectification is performed with a parametric extension of the now popular rectified linear units (ReLUs), whose parameters can be tuned for the detection of target object classes. This enables a number of functional enhancements over neural network models that lack a connection to saliency, including optimal feature denoising mechanisms for recognition, modulation of saliency responses by the discriminant power of the underlying features, and the ability to detect both feature presence and absence. In either implementation, each layer has a precise statistical interpretation, and all parameters are tuned by statistical learning. Each saliency detection layer learns more discriminant saliency templates than its predecessors and higher layers have larger pooling fields. This enables the HDSN to simultaneously achieve high selectivity to target object classes and invariance. The performance of the network in saliency and object recognition tasks is compared to those of models from the biological and computer vision literatures. This demonstrates benefits for all the functional enhancements of the HDSN, the class tuning inherent to discriminant saliency, and saliency layers based on templates of increasing target selectivity and invariance. Altogether, these experiments suggest that there are non-trivial benefits in integrating attention and recognition.

Investigation on trophic state index by artificial neural networks (case study: Dez Dam of Iran)

NASA Astrophysics Data System (ADS)

Saghi, H.; Karimi, L.; Javid, A. H.

2015-06-01

Dam construction and surface runoff control is one of the most common approaches for water-needs supply of human societies. However, the increasing development of social activities and hence the subsequent increase in environmental pollutants leads to deterioration of water quality in dam reservoirs and eutrophication process could be intensified. So, the water quality of reservoirs is now one of the key factors in operation and water quality management of reservoirs. Hence, maintaining the quality of the stored water and identification and examination of changes along time has been a constant concern of humans that involves the water authorities. Traditionally, empirical trophic state indices of dam reservoirs often defined based on changes in concentration of effective factors (nutrients) and its consequences (increase in chlorophyll a), have been used as an efficient tool in the definition of dam reservoirs quality. In recent years, modeling techniques such as artificial neural networks have enhanced the prediction capability and the accuracy of these studies. In this study, artificial neural networks have been applied to analyze eutrophication process in the Dez Dam reservoir in Iran. In this paper, feed forward neural network with one input layer, one hidden layer and one output layer was applied using MATLAB neural network toolbox for trophic state index (TSI) analysis in the Dez Dam reservoir. The input data of this network are effective parameters in the eutrophication: nitrogen cycle parameters and phosphorous cycle parameters and parameters that will be changed by eutrophication: Chl a, SD, DO and the output data is TSI. Based on the results from estimation of modified Carlson trophic state index, Dez Dam reservoir is considered to be eutrophic in the early July to mid-November and would be mesotrophic with decrease in temperature. Therefore, a decrease in water quality of the dam reservoir during the warm seasons is expectable. The results indicated that artificial neural network (ANN) is a suitable tool for quality modeling of reservoir of dam and increment and decrement of nutrients in trend of eutrophication. Therefore, ANN is a suitable tool for quality modeling of reservoir of dam.

Training strategy for convolutional neural networks in pedestrian gender classification

NASA Astrophysics Data System (ADS)

Ng, Choon-Boon; Tay, Yong-Haur; Goi, Bok-Min

2017-06-01

In this work, we studied a strategy for training a convolutional neural network in pedestrian gender classification with limited amount of labeled training data. Unsupervised learning by k-means clustering on pedestrian images was used to learn the filters to initialize the first layer of the network. As a form of pre-training, supervised learning for the related task of pedestrian classification was performed. Finally, the network was fine-tuned for gender classification. We found that this strategy improved the network's generalization ability in gender classification, achieving better test results when compared to random weights initialization and slightly more beneficial than merely initializing the first layer filters by unsupervised learning. This shows that unsupervised learning followed by pre-training with pedestrian images is an effective strategy to learn useful features for pedestrian gender classification.

Transfer Learning with Convolutional Neural Networks for Classification of Abdominal Ultrasound Images.

PubMed

Cheng, Phillip M; Malhi, Harshawn S

2017-04-01

The purpose of this study is to evaluate transfer learning with deep convolutional neural networks for the classification of abdominal ultrasound images. Grayscale images from 185 consecutive clinical abdominal ultrasound studies were categorized into 11 categories based on the text annotation specified by the technologist for the image. Cropped images were rescaled to 256 × 256 resolution and randomized, with 4094 images from 136 studies constituting the training set, and 1423 images from 49 studies constituting the test set. The fully connected layers of two convolutional neural networks based on CaffeNet and VGGNet, previously trained on the 2012 Large Scale Visual Recognition Challenge data set, were retrained on the training set. Weights in the convolutional layers of each network were frozen to serve as fixed feature extractors. Accuracy on the test set was evaluated for each network. A radiologist experienced in abdominal ultrasound also independently classified the images in the test set into the same 11 categories. The CaffeNet network classified 77.3% of the test set images accurately (1100/1423 images), with a top-2 accuracy of 90.4% (1287/1423 images). The larger VGGNet network classified 77.9% of the test set accurately (1109/1423 images), with a top-2 accuracy of VGGNet was 89.7% (1276/1423 images). The radiologist classified 71.7% of the test set images correctly (1020/1423 images). The differences in classification accuracies between both neural networks and the radiologist were statistically significant (p < 0.001). The results demonstrate that transfer learning with convolutional neural networks may be used to construct effective classifiers for abdominal ultrasound images.

Parameter diagnostics of phases and phase transition learning by neural networks

NASA Astrophysics Data System (ADS)

Suchsland, Philippe; Wessel, Stefan

2018-05-01

We present an analysis of neural network-based machine learning schemes for phases and phase transitions in theoretical condensed matter research, focusing on neural networks with a single hidden layer. Such shallow neural networks were previously found to be efficient in classifying phases and locating phase transitions of various basic model systems. In order to rationalize the emergence of the classification process and for identifying any underlying physical quantities, it is feasible to examine the weight matrices and the convolutional filter kernels that result from the learning process of such shallow networks. Furthermore, we demonstrate how the learning-by-confusing scheme can be used, in combination with a simple threshold-value classification method, to diagnose the learning parameters of neural networks. In particular, we study the classification process of both fully-connected and convolutional neural networks for the two-dimensional Ising model with extended domain wall configurations included in the low-temperature regime. Moreover, we consider the two-dimensional XY model and contrast the performance of the learning-by-confusing scheme and convolutional neural networks trained on bare spin configurations to the case of preprocessed samples with respect to vortex configurations. We discuss these findings in relation to similar recent investigations and possible further applications.

Artificial and Bayesian Neural Networks

PubMed

Korhani Kangi, Azam; Bahrampour, Abbas

2018-02-26

Introduction and purpose: In recent years the use of neural networks without any premises for investigation of prognosis in analyzing survival data has increased. Artificial neural networks (ANN) use small processors with a continuous network to solve problems inspired by the human brain. Bayesian neural networks (BNN) constitute a neural-based approach to modeling and non-linearization of complex issues using special algorithms and statistical methods. Gastric cancer incidence is the first and third ranking for men and women in Iran, respectively. The aim of the present study was to assess the value of an artificial neural network and a Bayesian neural network for modeling and predicting of probability of gastric cancer patient death. Materials and Methods: In this study, we used information on 339 patients aged from 20 to 90 years old with positive gastric cancer, referred to Afzalipoor and Shahid Bahonar Hospitals in Kerman City from 2001 to 2015. The three layers perceptron neural network (ANN) and the Bayesian neural network (BNN) were used for predicting the probability of mortality using the available data. To investigate differences between the models, sensitivity, specificity, accuracy and the area under receiver operating characteristic curves (AUROCs) were generated. Results: In this study, the sensitivity and specificity of the artificial neural network and Bayesian neural network models were 0.882, 0.903 and 0.954, 0.909, respectively. Prediction accuracy and the area under curve ROC for the two models were 0.891, 0.944 and 0.935, 0.961. The age at diagnosis of gastric cancer was most important for predicting survival, followed by tumor grade, morphology, gender, smoking history, opium consumption, receiving chemotherapy, presence of metastasis, tumor stage, receiving radiotherapy, and being resident in a village. Conclusion: The findings of the present study indicated that the Bayesian neural network is preferable to an artificial neural network for predicting survival of gastric cancer patients in Iran. Creative Commons Attribution License

Development of a neural network technique for KSTAR Thomson scattering diagnostics.

PubMed

Lee, Seung Hun; Lee, J H; Yamada, I; Park, Jae Sun

2016-11-01

Neural networks provide powerful approaches of dealing with nonlinear data and have been successfully applied to fusion plasma diagnostics and control systems. Controlling tokamak plasmas in real time is essential to measure the plasma parameters in situ. However, the χ 2 method traditionally used in Thomson scattering diagnostics hampers real-time measurement due to the complexity of the calculations involved. In this study, we applied a neural network approach to Thomson scattering diagnostics in order to calculate the electron temperature, comparing the results to those obtained with the χ 2 method. The best results were obtained for 10 3 training cycles and eight nodes in the hidden layer. Our neural network approach shows good agreement with the χ 2 method and performs the calculation twenty times faster.

Detection of Coal Fires: A Case Study Conducted on Indian Coal Seams Using Neural Network and Particle Swarm Optimization

NASA Astrophysics Data System (ADS)

Singh, B. B.

2016-12-01

India produces majority of its electricity from coal but a huge quantity of coal burns every day due to coal fires and also poses a threat to the environment as severe pollutants. In the present study we had demonstrated the usage of Neural Network based approach with an integrated Particle Swarm Optimization (PSO) inversion technique. The Self Potential (SP) data set is used for the early detection of coal fires. The study was conducted over the East Basuria colliery, Jharia Coal Field, Jharkhand, India. The causative source was modelled as an inclined sheet like anomaly and the synthetic data was generated. Neural Network scheme consists of an input layer, hidden layers and an output layer. The input layer corresponds to the SP data and the output layer is the estimated depth of the coal fire. A synthetic dataset was modelled with some of the known parameters such as depth, conductivity, inclination angle, half width etc. associated with causative body and gives a very low misfit error of 0.0032%. Therefore, the method was found accurate in predicting the depth of the source body. The technique was applied to the real data set and the model was trained until a very good correlation of determination `R2' value of 0.98 is obtained. The depth of the source body was found to be 12.34m with a misfit error percentage of 0.242%. The inversion results were compared with the lithologs obtained from a nearby well which corresponds to the L3 coal seam. The depth of the coal fire had exactly matched with the half width of the anomaly which suggests that the fire is widely spread. The inclination angle of the anomaly was 135.510 which resembles the development of the geometrically complex fracture planes. These fractures may be developed due to anisotropic weakness of the ground which acts as passage for the air. As a result coal fires spreads along these fracture planes. The results obtained from the Neural Network was compared with PSO inversion results and were found in complete agreement. PSO technique had already been found a well-established technique to model SP anomalies. Therefore for successful control and mitigation, SP surveys coupled with Neural Network and PSO technique proves to be novel and economical approach along with other existing geophysical techniques. Keywords: PSO, Coal fire, Self-Potential, Inversion, Neural Network

Inversion of surface parameters using fast learning neural networks

NASA Technical Reports Server (NTRS)

Dawson, M. S.; Olvera, J.; Fung, A. K.; Manry, M. T.

1992-01-01

A neural network approach to the inversion of surface scattering parameters is presented. Simulated data sets based on a surface scattering model are used so that the data may be viewed as taken from a completely known randomly rough surface. The fast learning (FL) neural network and a multilayer perceptron (MLP) trained with backpropagation learning (BP network) are tested on the simulated backscattering data. The RMS error of training the FL network is found to be less than one half the error of the BP network while requiring one to two orders of magnitude less CPU time. When applied to inversion of parameters from a statistically rough surface, the FL method is successful at recovering the surface permittivity, the surface correlation length, and the RMS surface height in less time and with less error than the BP network. Further applications of the FL neural network to the inversion of parameters from backscatter measurements of an inhomogeneous layer above a half space are shown.

Neural Network Modeling for Gallium Arsenide IC Fabrication Process and Device Characteristics.

NASA Astrophysics Data System (ADS)

Creech, Gregory Lee, I.

This dissertation presents research focused on the utilization of neurocomputing technology to achieve enhanced yield and effective yield prediction in integrated circuit (IC) manufacturing. Artificial neural networks are employed to model complex relationships between material and device characteristics at critical stages of the semiconductor fabrication process. Whole wafer testing was performed on the starting substrate material and during wafer processing at four critical steps: Ohmic or Post-Contact, Post-Recess, Post-Gate and Final, i.e., at completion of fabrication. Measurements taken and subsequently used in modeling include, among others, doping concentrations, layer thicknesses, planar geometries, layer-to-layer alignments, resistivities, device voltages, and currents. The neural network architecture used in this research is the multilayer perceptron neural network (MLPNN). The MLPNN is trained in the supervised mode using the generalized delta learning rule. It has one hidden layer and uses continuous perceptrons. The research focuses on a number of different aspects. First is the development of inter-process stage models. Intermediate process stage models are created in a progressive fashion. Measurements of material and process/device characteristics taken at a specific processing stage and any previous stages are used as input to the model of the next processing stage characteristics. As the wafer moves through the fabrication process, measurements taken at all previous processing stages are used as input to each subsequent process stage model. Secondly, the development of neural network models for the estimation of IC parametric yield is demonstrated. Measurements of material and/or device characteristics taken at earlier fabrication stages are used to develop models of the final DC parameters. These characteristics are computed with the developed models and compared to acceptance windows to estimate the parametric yield. A sensitivity analysis is performed on the models developed during this yield estimation effort. This is accomplished by analyzing the total disturbance of network outputs due to perturbed inputs. When an input characteristic bears no, or little, statistical or deterministic relationship to the output characteristics, it can be removed as an input. Finally, neural network models are developed in the inverse direction. Characteristics measured after the final processing step are used as the input to model critical in-process characteristics. The modeled characteristics are used for whole wafer mapping and its statistical characterization. It is shown that this characterization can be accomplished with minimal in-process testing. The concepts and methodologies used in the development of the neural network models are presented. The modeling results are provided and compared to the actual measured values of each characteristic. An in-depth discussion of these results and ideas for future research are presented.

Sensitivity of feedforward neural networks to weight errors

NASA Technical Reports Server (NTRS)

Stevenson, Maryhelen; Widrow, Bernard; Winter, Rodney

1990-01-01

An analysis is made of the sensitivity of feedforward layered networks of Adaline elements (threshold logic units) to weight errors. An approximation is derived which expresses the probability of error for an output neuron of a large network (a network with many neurons per layer) as a function of the percentage change in the weights. As would be expected, the probability of error increases with the number of layers in the network and with the percentage change in the weights. The probability of error is essentially independent of the number of weights per neuron and of the number of neurons per layer, as long as these numbers are large (on the order of 100 or more).

Germ layers, the neural crest and emergent organization in development and evolution.

PubMed

Hall, Brian K

2018-04-10

Discovered in chick embryos by Wilhelm His in 1868 and named the neural crest by Arthur Milnes Marshall in 1879, the neural crest cells that arise from the neural folds have since been shown to differentiate into almost two dozen vertebrate cell types and to have played major roles in the evolution of such vertebrate features as bone, jaws, teeth, visceral (pharyngeal) arches, and sense organs. I discuss the discovery that ectodermal neural crest gave rise to mesenchyme and the controversy generated by that finding; the germ layer theory maintained that only mesoderm could give rise to mesenchyme. A second topic of discussion is germ layers (including the neural crest) as emergent levels of organization in animal development and evolution that facilitated major developmental and evolutionary change. The third topic is gene networks, gene co-option, and the evolution of gene-signaling pathways as key to developmental and evolutionary transitions associated with the origin and evolution of the neural crest and neural crest cells. © 2018 Wiley Periodicals, Inc.

KNT-artificial neural network model for flux prediction of ultrafiltration membrane producing drinking water.

PubMed

Oh, H K; Yu, M J; Gwon, E M; Koo, J Y; Kim, S G; Koizumi, A

2004-01-01

This paper describes the prediction of flux behavior in an ultrafiltration (UF) membrane system using a Kalman neuro training (KNT) network model. The experimental data was obtained from operating a pilot plant of hollow fiber UF membrane with groundwater for 7 months. The network was trained using operating conditions such as inlet pressure, filtration duration, and feed water quality parameters including turbidity, temperature and UV254. Pre-processing of raw data allowed the normalized input data to be used in sigmoid activation functions. A neural network architecture was structured by modifying the number of hidden layers, neurons and learning iterations. The structure of KNT-neural network with 3 layers and 5 neurons allowed a good prediction of permeate flux by 0.997 of correlation coefficient during the learning phase. Also the validity of the designed model was evaluated with other experimental data not used during the training phase and nonlinear flux behavior was accurately estimated with 0.999 of correlation coefficient and a lower error of prediction in the testing phase. This good flux prediction can provide preliminary criteria in membrane design and set up the proper cleaning cycle in membrane operation. The KNT-artificial neural network is also expected to predict the variation of transmembrane pressure during filtration cycles and can be applied to automation and control of full scale treatment plants.

Signal processing and neural network toolbox and its application to failure diagnosis and prognosis

NASA Astrophysics Data System (ADS)

Tu, Fang; Wen, Fang; Willett, Peter K.; Pattipati, Krishna R.; Jordan, Eric H.

2001-07-01

Many systems are comprised of components equipped with self-testing capability; however, if the system is complex involving feedback and the self-testing itself may occasionally be faulty, tracing faults to a single or multiple causes is difficult. Moreover, many sensors are incapable of reliable decision-making on their own. In such cases, a signal processing front-end that can match inference needs will be very helpful. The work is concerned with providing an object-oriented simulation environment for signal processing and neural network-based fault diagnosis and prognosis. In the toolbox, we implemented a wide range of spectral and statistical manipulation methods such as filters, harmonic analyzers, transient detectors, and multi-resolution decomposition to extract features for failure events from data collected by data sensors. Then we evaluated multiple learning paradigms for general classification, diagnosis and prognosis. The network models evaluated include Restricted Coulomb Energy (RCE) Neural Network, Learning Vector Quantization (LVQ), Decision Trees (C4.5), Fuzzy Adaptive Resonance Theory (FuzzyArtmap), Linear Discriminant Rule (LDR), Quadratic Discriminant Rule (QDR), Radial Basis Functions (RBF), Multiple Layer Perceptrons (MLP) and Single Layer Perceptrons (SLP). Validation techniques, such as N-fold cross-validation and bootstrap techniques, are employed for evaluating the robustness of network models. The trained networks are evaluated for their performance using test data on the basis of percent error rates obtained via cross-validation, time efficiency, generalization ability to unseen faults. Finally, the usage of neural networks for the prediction of residual life of turbine blades with thermal barrier coatings is described and the results are shown. The neural network toolbox has also been applied to fault diagnosis in mixed-signal circuits.

Deep Neural Network for Structural Prediction and Lane Detection in Traffic Scene.

PubMed

Li, Jun; Mei, Xue; Prokhorov, Danil; Tao, Dacheng

2017-03-01

Hierarchical neural networks have been shown to be effective in learning representative image features and recognizing object classes. However, most existing networks combine the low/middle level cues for classification without accounting for any spatial structures. For applications such as understanding a scene, how the visual cues are spatially distributed in an image becomes essential for successful analysis. This paper extends the framework of deep neural networks by accounting for the structural cues in the visual signals. In particular, two kinds of neural networks have been proposed. First, we develop a multitask deep convolutional network, which simultaneously detects the presence of the target and the geometric attributes (location and orientation) of the target with respect to the region of interest. Second, a recurrent neuron layer is adopted for structured visual detection. The recurrent neurons can deal with the spatial distribution of visible cues belonging to an object whose shape or structure is difficult to explicitly define. Both the networks are demonstrated by the practical task of detecting lane boundaries in traffic scenes. The multitask convolutional neural network provides auxiliary geometric information to help the subsequent modeling of the given lane structures. The recurrent neural network automatically detects lane boundaries, including those areas containing no marks, without any explicit prior knowledge or secondary modeling.

Network evolution induced by asynchronous stimuli through spike-timing-dependent plasticity.

PubMed

Yuan, Wu-Jie; Zhou, Jian-Fang; Zhou, Changsong

2013-01-01

In sensory neural system, external asynchronous stimuli play an important role in perceptual learning, associative memory and map development. However, the organization of structure and dynamics of neural networks induced by external asynchronous stimuli are not well understood. Spike-timing-dependent plasticity (STDP) is a typical synaptic plasticity that has been extensively found in the sensory systems and that has received much theoretical attention. This synaptic plasticity is highly sensitive to correlations between pre- and postsynaptic firings. Thus, STDP is expected to play an important role in response to external asynchronous stimuli, which can induce segregative pre- and postsynaptic firings. In this paper, we study the impact of external asynchronous stimuli on the organization of structure and dynamics of neural networks through STDP. We construct a two-dimensional spatial neural network model with local connectivity and sparseness, and use external currents to stimulate alternately on different spatial layers. The adopted external currents imposed alternately on spatial layers can be here regarded as external asynchronous stimuli. Through extensive numerical simulations, we focus on the effects of stimulus number and inter-stimulus timing on synaptic connecting weights and the property of propagation dynamics in the resulting network structure. Interestingly, the resulting feedforward structure induced by stimulus-dependent asynchronous firings and its propagation dynamics reflect both the underlying property of STDP. The results imply a possible important role of STDP in generating feedforward structure and collective propagation activity required for experience-dependent map plasticity in developing in vivo sensory pathways and cortices. The relevance of the results to cue-triggered recall of learned temporal sequences, an important cognitive function, is briefly discussed as well. Furthermore, this finding suggests a potential application for examining STDP by measuring neural population activity in a cultured neural network.

Neural Network and Response Surface Methodology for Rocket Engine Component Optimization

NASA Technical Reports Server (NTRS)

Vaidyanathan, Rajkumar; Papita, Nilay; Shyy, Wei; Tucker, P. Kevin; Griffin, Lisa W.; Haftka, Raphael; Fitz-Coy, Norman; McConnaughey, Helen (Technical Monitor)

2000-01-01

The goal of this work is to compare the performance of response surface methodology (RSM) and two types of neural networks (NN) to aid preliminary design of two rocket engine components. A data set of 45 training points and 20 test points obtained from a semi-empirical model based on three design variables is used for a shear coaxial injector element. Data for supersonic turbine design is based on six design variables, 76 training, data and 18 test data obtained from simplified aerodynamic analysis. Several RS and NN are first constructed using the training data. The test data are then employed to select the best RS or NN. Quadratic and cubic response surfaces. radial basis neural network (RBNN) and back-propagation neural network (BPNN) are compared. Two-layered RBNN are generated using two different training algorithms, namely solverbe and solverb. A two layered BPNN is generated with Tan-Sigmoid transfer function. Various issues related to the training of the neural networks are addressed including number of neurons, error goals, spread constants and the accuracy of different models in representing the design space. A search for the optimum design is carried out using a standard gradient-based optimization algorithm over the response surfaces represented by the polynomials and trained neural networks. Usually a cubic polynominal performs better than the quadratic polynomial but exceptions have been noticed. Among the NN choices, the RBNN designed using solverb yields more consistent performance for both engine components considered. The training of RBNN is easier as it requires linear regression. This coupled with the consistency in performance promise the possibility of it being used as an optimization strategy for engineering design problems.

Neural-network-designed pulse sequences for robust control of singlet-triplet qubits

NASA Astrophysics Data System (ADS)

Yang, Xu-Chen; Yung, Man-Hong; Wang, Xin

2018-04-01

Composite pulses are essential for universal manipulation of singlet-triplet spin qubits. In the absence of noise, they are required to perform arbitrary single-qubit operations due to the special control constraint of a singlet-triplet qubit, while in a noisy environment, more complicated sequences have been developed to dynamically correct the error. Tailoring these sequences typically requires numerically solving a set of nonlinear equations. Here we demonstrate that these pulse sequences can be generated by a well-trained, double-layer neural network. For sequences designed for the noise-free case, the trained neural network is capable of producing almost exactly the same pulses known in the literature. For more complicated noise-correcting sequences, the neural network produces pulses with slightly different line shapes, but the robustness against noises remains comparable. These results indicate that the neural network can be a judicious and powerful alternative to existing techniques in developing pulse sequences for universal fault-tolerant quantum computation.

Combined neural network/Phillips-Tikhonov approach to aerosol retrievals over land from the NASA Research Scanning Polarimeter

NASA Astrophysics Data System (ADS)

Di Noia, Antonio; Hasekamp, Otto P.; Wu, Lianghai; van Diedenhoven, Bastiaan; Cairns, Brian; Yorks, John E.

2017-11-01

In this paper, an algorithm for the retrieval of aerosol and land surface properties from airborne spectropolarimetric measurements - combining neural networks and an iterative scheme based on Phillips-Tikhonov regularization - is described. The algorithm - which is an extension of a scheme previously designed for ground-based retrievals - is applied to measurements from the Research Scanning Polarimeter (RSP) on board the NASA ER-2 aircraft. A neural network, trained on a large data set of synthetic measurements, is applied to perform aerosol retrievals from real RSP data, and the neural network retrievals are subsequently used as a first guess for the Phillips-Tikhonov retrieval. The resulting algorithm appears capable of accurately retrieving aerosol optical thickness, fine-mode effective radius and aerosol layer height from RSP data. Among the advantages of using a neural network as initial guess for an iterative algorithm are a decrease in processing time and an increase in the number of converging retrievals.

[Evaluation of eco-environmental quality based on artificial neural network and remote sensing techniques].

PubMed

Li, Hongyi; Shi, Zhou; Sha, Jinming; Cheng, Jieliang

2006-08-01

In the present study, vegetation, soil brightness, and moisture indices were extracted from Landsat ETM remote sensing image, heat indices were extracted from MODIS land surface temperature product, and climate index and other auxiliary geographical information were selected as the input of neural network. The remote sensing eco-environmental background value of standard interest region evaluated in situ was selected as the output of neural network, and the back propagation (BP) neural network prediction model containing three layers was designed. The network was trained, and the remote sensing eco-environmental background value of Fuzhou in China was predicted by using software MATLAB. The class mapping of remote sensing eco-environmental background values based on evaluation standard showed that the total classification accuracy was 87. 8%. The method with a scheme of prediction first and classification then could provide acceptable results in accord with the regional eco-environment types.

Dynamic boundary layer based neural network quasi-sliding mode control for soft touching down on asteroid

NASA Astrophysics Data System (ADS)

Liu, Xiaosong; Shan, Zebiao; Li, Yuanchun

2017-04-01

Pinpoint landing is a critical step in some asteroid exploring missions. This paper is concerned with the descent trajectory control for soft touching down on a small irregularly-shaped asteroid. A dynamic boundary layer based neural network quasi-sliding mode control law is proposed to track a desired descending path. The asteroid's gravitational acceleration acting on the spacecraft is described by the polyhedron method. Considering the presence of input constraint and unmodeled acceleration, the dynamic equation of relative motion is presented first. The desired descending path is planned using cubic polynomial method, and a collision detection algorithm is designed. To perform trajectory tracking, a neural network sliding mode control law is given first, where the sliding mode control is used to ensure the convergence of system states. Two radial basis function neural networks (RBFNNs) are respectively used as an approximator for the unmodeled term and a compensator for the difference between the actual control input with magnitude constraint and nominal control. To improve the chattering induced by the traditional sliding mode control and guarantee the reachability of the system, a specific saturation function with dynamic boundary layer is proposed to replace the sign function in the preceding control law. Through the Lyapunov approach, the reachability condition of the control system is given. The improved control law can guarantee the system state move within a gradually shrinking quasi-sliding mode band. Numerical simulation results demonstrate the effectiveness of the proposed control strategy.

Yarn-dyed fabric defect classification based on convolutional neural network

NASA Astrophysics Data System (ADS)

Jing, Junfeng; Dong, Amei; Li, Pengfei; Zhang, Kaibing

2017-09-01

Considering that manual inspection of the yarn-dyed fabric can be time consuming and inefficient, we propose a yarn-dyed fabric defect classification method by using a convolutional neural network (CNN) based on a modified AlexNet. CNN shows powerful ability in performing feature extraction and fusion by simulating the learning mechanism of human brain. The local response normalization layers in AlexNet are replaced by the batch normalization layers, which can enhance both the computational efficiency and classification accuracy. In the training process of the network, the characteristics of the defect are extracted step by step and the essential features of the image can be obtained from the fusion of the edge details with several convolution operations. Then the max-pooling layers, the dropout layers, and the fully connected layers are employed in the classification model to reduce the computation cost and extract more precise features of the defective fabric. Finally, the results of the defect classification are predicted by the softmax function. The experimental results show promising performance with an acceptable average classification rate and strong robustness on yarn-dyed fabric defect classification.

Yarn-dyed fabric defect classification based on convolutional neural network

NASA Astrophysics Data System (ADS)

Jing, Junfeng; Dong, Amei; Li, Pengfei

2017-07-01

Considering that the manual inspection of the yarn-dyed fabric can be time consuming and less efficient, a convolutional neural network (CNN) solution based on the modified AlexNet structure for the classification of the yarn-dyed fabric defect is proposed. CNN has powerful ability of feature extraction and feature fusion which can simulate the learning mechanism of the human brain. In order to enhance computational efficiency and detection accuracy, the local response normalization (LRN) layers in AlexNet are replaced by the batch normalization (BN) layers. In the process of the network training, through several convolution operations, the characteristics of the image are extracted step by step, and the essential features of the image can be obtained from the edge features. And the max pooling layers, the dropout layers, the fully connected layers are also employed in the classification model to reduce the computation cost and acquire more precise features of fabric defect. Finally, the results of the defect classification are predicted by the softmax function. The experimental results show the capability of defect classification via the modified Alexnet model and indicate its robustness.

Task-specific feature extraction and classification of fMRI volumes using a deep neural network initialized with a deep belief network: Evaluation using sensorimotor tasks

PubMed Central

Jang, Hojin; Plis, Sergey M.; Calhoun, Vince D.; Lee, Jong-Hwan

2016-01-01

Feedforward deep neural networks (DNN), artificial neural networks with multiple hidden layers, have recently demonstrated a record-breaking performance in multiple areas of applications in computer vision and speech processing. Following the success, DNNs have been applied to neuroimaging modalities including functional/structural magnetic resonance imaging (MRI) and positron-emission tomography data. However, no study has explicitly applied DNNs to 3D whole-brain fMRI volumes and thereby extracted hidden volumetric representations of fMRI that are discriminative for a task performed as the fMRI volume was acquired. Our study applied fully connected feedforward DNN to fMRI volumes collected in four sensorimotor tasks (i.e., left-hand clenching, right-hand clenching, auditory attention, and visual stimulus) undertaken by 12 healthy participants. Using a leave-one-subject-out cross-validation scheme, a restricted Boltzmann machine-based deep belief network was pretrained and used to initialize weights of the DNN. The pretrained DNN was fine-tuned while systematically controlling weight-sparsity levels across hidden layers. Optimal weight-sparsity levels were determined from a minimum validation error rate of fMRI volume classification. Minimum error rates (mean ± standard deviation; %) of 6.9 (± 3.8) were obtained from the three-layer DNN with the sparsest condition of weights across the three hidden layers. These error rates were even lower than the error rates from the single-layer network (9.4 ± 4.6) and the two-layer network (7.4 ± 4.1). The estimated DNN weights showed spatial patterns that are remarkably task-specific, particularly in the higher layers. The output values of the third hidden layer represented distinct patterns/codes of the 3D whole-brain fMRI volume and encoded the information of the tasks as evaluated from representational similarity analysis. Our reported findings show the ability of the DNN to classify a single fMRI volume based on the extraction of hidden representations of fMRI volumes associated with tasks across multiple hidden layers. Our study may be beneficial to the automatic classification/diagnosis of neuropsychiatric and neurological diseases and prediction of disease severity and recovery in (pre-) clinical settings using fMRI volumes without requiring an estimation of activation patterns or ad hoc statistical evaluation. PMID:27079534

Task-specific feature extraction and classification of fMRI volumes using a deep neural network initialized with a deep belief network: Evaluation using sensorimotor tasks.

PubMed

Jang, Hojin; Plis, Sergey M; Calhoun, Vince D; Lee, Jong-Hwan

2017-01-15

Feedforward deep neural networks (DNNs), artificial neural networks with multiple hidden layers, have recently demonstrated a record-breaking performance in multiple areas of applications in computer vision and speech processing. Following the success, DNNs have been applied to neuroimaging modalities including functional/structural magnetic resonance imaging (MRI) and positron-emission tomography data. However, no study has explicitly applied DNNs to 3D whole-brain fMRI volumes and thereby extracted hidden volumetric representations of fMRI that are discriminative for a task performed as the fMRI volume was acquired. Our study applied fully connected feedforward DNN to fMRI volumes collected in four sensorimotor tasks (i.e., left-hand clenching, right-hand clenching, auditory attention, and visual stimulus) undertaken by 12 healthy participants. Using a leave-one-subject-out cross-validation scheme, a restricted Boltzmann machine-based deep belief network was pretrained and used to initialize weights of the DNN. The pretrained DNN was fine-tuned while systematically controlling weight-sparsity levels across hidden layers. Optimal weight-sparsity levels were determined from a minimum validation error rate of fMRI volume classification. Minimum error rates (mean±standard deviation; %) of 6.9 (±3.8) were obtained from the three-layer DNN with the sparsest condition of weights across the three hidden layers. These error rates were even lower than the error rates from the single-layer network (9.4±4.6) and the two-layer network (7.4±4.1). The estimated DNN weights showed spatial patterns that are remarkably task-specific, particularly in the higher layers. The output values of the third hidden layer represented distinct patterns/codes of the 3D whole-brain fMRI volume and encoded the information of the tasks as evaluated from representational similarity analysis. Our reported findings show the ability of the DNN to classify a single fMRI volume based on the extraction of hidden representations of fMRI volumes associated with tasks across multiple hidden layers. Our study may be beneficial to the automatic classification/diagnosis of neuropsychiatric and neurological diseases and prediction of disease severity and recovery in (pre-) clinical settings using fMRI volumes without requiring an estimation of activation patterns or ad hoc statistical evaluation. Copyright © 2016 Elsevier Inc. All rights reserved.

Design of Neural Networks for Fast Convergence and Accuracy

NASA Technical Reports Server (NTRS)

Maghami, Peiman G.; Sparks, Dean W., Jr.

1998-01-01

A novel procedure for the design and training of artificial neural networks, used for rapid and efficient controls and dynamics design and analysis for flexible space systems, has been developed. Artificial neural networks are employed to provide a means of evaluating the impact of design changes rapidly. Specifically, two-layer feedforward neural networks are designed to approximate the functional relationship between the component spacecraft design changes and measures of its performance. A training algorithm, based on statistical sampling theory, is presented, which guarantees that the trained networks provide a designer-specified degree of accuracy in mapping the functional relationship. Within each iteration of this statistical-based algorithm, a sequential design algorithm is used for the design and training of the feedforward network to provide rapid convergence to the network goals. Here, at each sequence a new network is trained to minimize the error of previous network. The design algorithm attempts to avoid the local minima phenomenon that hampers the traditional network training. A numerical example is performed on a spacecraft application in order to demonstrate the feasibility of the proposed approach.

Multilayer perceptron architecture optimization using parallel computing techniques.

PubMed

Castro, Wilson; Oblitas, Jimy; Santa-Cruz, Roberto; Avila-George, Himer

2017-01-01

The objective of this research was to develop a methodology for optimizing multilayer-perceptron-type neural networks by evaluating the effects of three neural architecture parameters, namely, number of hidden layers (HL), neurons per hidden layer (NHL), and activation function type (AF), on the sum of squares error (SSE). The data for the study were obtained from quality parameters (physicochemical and microbiological) of milk samples. Architectures or combinations were organized in groups (G1, G2, and G3) generated upon interspersing one, two, and three layers. Within each group, the networks had three neurons in the input layer, six neurons in the output layer, three to twenty-seven NHL, and three AF (tan-sig, log-sig, and linear) types. The number of architectures was determined using three factorial-type experimental designs, which reached 63, 2 187, and 50 049 combinations for G1, G2 and G3, respectively. Using MATLAB 2015a, a logical sequence was designed and implemented for constructing, training, and evaluating multilayer-perceptron-type neural networks using parallel computing techniques. The results show that HL and NHL have a statistically relevant effect on SSE, and from two hidden layers, AF also has a significant effect; thus, both AF and NHL can be evaluated to determine the optimal combination per group. Moreover, in the three study groups, it is observed that there is an inverse relationship between the number of processors and the total optimization time.

Multilayer perceptron architecture optimization using parallel computing techniques

PubMed Central

Castro, Wilson; Oblitas, Jimy; Santa-Cruz, Roberto; Avila-George, Himer

2017-01-01

The objective of this research was to develop a methodology for optimizing multilayer-perceptron-type neural networks by evaluating the effects of three neural architecture parameters, namely, number of hidden layers (HL), neurons per hidden layer (NHL), and activation function type (AF), on the sum of squares error (SSE). The data for the study were obtained from quality parameters (physicochemical and microbiological) of milk samples. Architectures or combinations were organized in groups (G1, G2, and G3) generated upon interspersing one, two, and three layers. Within each group, the networks had three neurons in the input layer, six neurons in the output layer, three to twenty-seven NHL, and three AF (tan-sig, log-sig, and linear) types. The number of architectures was determined using three factorial-type experimental designs, which reached 63, 2 187, and 50 049 combinations for G1, G2 and G3, respectively. Using MATLAB 2015a, a logical sequence was designed and implemented for constructing, training, and evaluating multilayer-perceptron-type neural networks using parallel computing techniques. The results show that HL and NHL have a statistically relevant effect on SSE, and from two hidden layers, AF also has a significant effect; thus, both AF and NHL can be evaluated to determine the optimal combination per group. Moreover, in the three study groups, it is observed that there is an inverse relationship between the number of processors and the total optimization time. PMID:29236744

Prediction of hearing loss among the noise-exposed workers in a steel factory using artificial intelligence approach.

PubMed

Aliabadi, Mohsen; Farhadian, Maryam; Darvishi, Ebrahim

2015-08-01

Prediction of hearing loss in noisy workplaces is considered to be an important aspect of hearing conservation program. Artificial intelligence, as a new approach, can be used to predict the complex phenomenon such as hearing loss. Using artificial neural networks, this study aims to present an empirical model for the prediction of the hearing loss threshold among noise-exposed workers. Two hundred and ten workers employed in a steel factory were chosen, and their occupational exposure histories were collected. To determine the hearing loss threshold, the audiometric test was carried out using a calibrated audiometer. The personal noise exposure was also measured using a noise dosimeter in the workstations of workers. Finally, data obtained five variables, which can influence the hearing loss, were used for the development of the prediction model. Multilayer feed-forward neural networks with different structures were developed using MATLAB software. Neural network structures had one hidden layer with the number of neurons being approximately between 5 and 15 neurons. The best developed neural networks with one hidden layer and ten neurons could accurately predict the hearing loss threshold with RMSE = 2.6 dB and R(2) = 0.89. The results also confirmed that neural networks could provide more accurate predictions than multiple regressions. Since occupational hearing loss is frequently non-curable, results of accurate prediction can be used by occupational health experts to modify and improve noise exposure conditions.

Cognitive Processing Hardware Elements

DTIC Science & Technology

2005-01-31

characters. Results will be presented below. 1 4. Recognition of human faces. There are many other possible applications such as facial recognition and...For the experiments in facial recognition , we have used a 3-layer autoassociative neural network having the following specifications: "* The input...using the facial recognition system described in the section above as an example. This system uses an autoassociative neural network containing over 10

Nonparametric Representations for Integrated Inference, Control, and Sensing

DTIC Science & Technology

2015-10-01

Learning (ICML), 2013. [20] Jeff Donahue, Yangqing Jia, Oriol Vinyals, Judy Hoffman, Ning Zhang, Eric Tzeng, and Trevor Darrell. DeCAF: A deep ...unlimited. Multi-layer feature learning “SuperVision” Convolutional Neural Network (CNN) ImageNet Classification with Deep Convolutional Neural Networks...to develop a new framework for autonomous operations that will extend the state of the art in distributed learning and modeling from data, and

Characterization of Radar Signals Using Neural Networks

DTIC Science & Technology

1990-12-01

e***e*e*eeeeeeeeeeeesseeeeeese*eee*e*e************s /* Function Name: load.input.ptterns Number: 4.1 /* Description: This function determines wether ...XSE.last.layer Number: 8.5 */ /* Description: The function determines wether to backpropate the *f /* parameter by the sigmoidal or linear update...Sigmoidal Function," Mathematics of Control, Signals and Systems, 2:303-314 (March 1989). 6. Dayhoff, Judith E. Neural Network Architectures. New York: Van

Machine Learning Technique to Find Quantum Many-Body Ground States of Bosons on a Lattice

NASA Astrophysics Data System (ADS)

Saito, Hiroki; Kato, Masaya

2018-01-01

We have developed a variational method to obtain many-body ground states of the Bose-Hubbard model using feedforward artificial neural networks. A fully connected network with a single hidden layer works better than a fully connected network with multiple hidden layers, and a multilayer convolutional network is more efficient than a fully connected network. AdaGrad and Adam are optimization methods that work well. Moreover, we show that many-body ground states with different numbers of particles can be generated by a single network.

Hybrid neural network for density limit disruption prediction and avoidance on J-TEXT tokamak

NASA Astrophysics Data System (ADS)

Zheng, W.; Hu, F. R.; Zhang, M.; Chen, Z. Y.; Zhao, X. Q.; Wang, X. L.; Shi, P.; Zhang, X. L.; Zhang, X. Q.; Zhou, Y. N.; Wei, Y. N.; Pan, Y.; J-TEXT team

2018-05-01

Increasing the plasma density is one of the key methods in achieving an efficient fusion reaction. High-density operation is one of the hot topics in tokamak plasmas. Density limit disruptions remain an important issue for safe operation. An effective density limit disruption prediction and avoidance system is the key to avoid density limit disruptions for long pulse steady state operations. An artificial neural network has been developed for the prediction of density limit disruptions on the J-TEXT tokamak. The neural network has been improved from a simple multi-layer design to a hybrid two-stage structure. The first stage is a custom network which uses time series diagnostics as inputs to predict plasma density, and the second stage is a three-layer feedforward neural network to predict the probability of density limit disruptions. It is found that hybrid neural network structure, combined with radiation profile information as an input can significantly improve the prediction performance, especially the average warning time ({{T}warn} ). In particular, the {{T}warn} is eight times better than that in previous work (Wang et al 2016 Plasma Phys. Control. Fusion 58 055014) (from 5 ms to 40 ms). The success rate for density limit disruptive shots is above 90%, while, the false alarm rate for other shots is below 10%. Based on the density limit disruption prediction system and the real-time density feedback control system, the on-line density limit disruption avoidance system has been implemented on the J-TEXT tokamak.

A microsensor array for quantification of lubricant contaminants using a back propagation artificial neural network

NASA Astrophysics Data System (ADS)

Zhu, Xiaoliang; Du, Li; Liu, Bendong; Zhe, Jiang

2016-06-01

We present a method based on an electrochemical sensor array and a back propagation artificial neural network for detection and quantification of four properties of lubrication oil, namely water (0, 500 ppm, 1000 ppm), total acid number (TAN) (13.1, 13.7, 14.4, 15.6 mg KOH g-1), soot (0, 1%, 2%, 3%) and sulfur content (1.3%, 1.37%, 1.44%, 1.51%). The sensor array, consisting of four micromachined electrochemical sensors, detects the four properties with overlapping sensitivities. A total set of 36 oil samples containing mixtures of water, soot, and sulfuric acid with different concentrations were prepared for testing. The sensor array’s responses were then divided to three sets: training sets (80% data), validation sets (10%) and testing sets (10%). Several back propagation artificial neural network architectures were trained with the training and validation sets; one architecture with four input neurons, 50 and 5 neurons in the first and second hidden layer, and four neurons in the output layer was selected. The selected neural network was then tested using the four sets of testing data (10%). Test results demonstrated that the developed artificial neural network is able to quantitatively determine the four lubrication properties (water, TAN, soot, and sulfur content) with a maximum prediction error of 18.8%, 6.0%, 6.7%, and 5.4%, respectively, indicting a good match between the target and predicted values. With the developed network, the sensor array could be potentially used for online lubricant oil condition monitoring.

Development of a neural network technique for KSTAR Thomson scattering diagnostics

DOE Office of Scientific and Technical Information (OSTI.GOV)

Lee, Seung Hun, E-mail: leesh81@nfri.re.kr; Lee, J. H.; Yamada, I.

Neural networks provide powerful approaches of dealing with nonlinear data and have been successfully applied to fusion plasma diagnostics and control systems. Controlling tokamak plasmas in real time is essential to measure the plasma parameters in situ. However, the χ{sup 2} method traditionally used in Thomson scattering diagnostics hampers real-time measurement due to the complexity of the calculations involved. In this study, we applied a neural network approach to Thomson scattering diagnostics in order to calculate the electron temperature, comparing the results to those obtained with the χ{sup 2} method. The best results were obtained for 10{sup 3} training cyclesmore » and eight nodes in the hidden layer. Our neural network approach shows good agreement with the χ{sup 2} method and performs the calculation twenty times faster.« less

Estimation of sleep stages by an artificial neural network employing EEG, EMG and EOG.

PubMed

Tagluk, M Emin; Sezgin, Necmettin; Akin, Mehmet

2010-08-01

Analysis and classification of sleep stages is essential in sleep research. In this particular study, an alternative system which estimates sleep stages of human being through a multi-layer neural network (NN) that simultaneously employs EEG, EMG and EOG. The data were recorded through polisomnography device for 7 h for each subject. These collective variant data were first grouped by an expert physician and the software of polisomnography, and then used for training and testing the proposed Artificial Neural Network (ANN). A good scoring was attained through the trained ANN, so it may be put into use in clinics where lacks of specialist physicians.

Improved head direction command classification using an optimised Bayesian neural network.

PubMed

Nguyen, Son T; Nguyen, Hung T; Taylor, Philip B; Middleton, James

2006-01-01

Assistive technologies have recently emerged to improve the quality of life of severely disabled people by enhancing their independence in daily activities. Since many of those individuals have limited or non-existing control from the neck downward, alternative hands-free input modalities have become very important for these people to access assistive devices. In hands-free control, head movement has been proved to be a very effective user interface as it can provide a comfortable, reliable and natural way to access the device. Recently, neural networks have been shown to be useful not only for real-time pattern recognition but also for creating user-adaptive models. Since multi-layer perceptron neural networks trained using standard back-propagation may cause poor generalisation, the Bayesian technique has been proposed to improve the generalisation and robustness of these networks. This paper describes the use of Bayesian neural networks in developing a hands-free wheelchair control system. The experimental results show that with the optimised architecture, classification Bayesian neural networks can detect head commands of wheelchair users accurately irrespective to their levels of injuries.

Estimation of effective connectivity using multi-layer perceptron artificial neural network.

PubMed

Talebi, Nasibeh; Nasrabadi, Ali Motie; Mohammad-Rezazadeh, Iman

2018-02-01

Studies on interactions between brain regions estimate effective connectivity, (usually) based on the causality inferences made on the basis of temporal precedence. In this study, the causal relationship is modeled by a multi-layer perceptron feed-forward artificial neural network, because of the ANN's ability to generate appropriate input-output mapping and to learn from training examples without the need of detailed knowledge of the underlying system. At any time instant, the past samples of data are placed in the network input, and the subsequent values are predicted at its output. To estimate the strength of interactions, the measure of " Causality coefficient " is defined based on the network structure, the connecting weights and the parameters of hidden layer activation function. Simulation analysis demonstrates that the method, called "CREANN" (Causal Relationship Estimation by Artificial Neural Network), can estimate time-invariant and time-varying effective connectivity in terms of MVAR coefficients. The method shows robustness with respect to noise level of data. Furthermore, the estimations are not significantly influenced by the model order (considered time-lag), and the different initial conditions (initial random weights and parameters of the network). CREANN is also applied to EEG data collected during a memory recognition task. The results implicate that it can show changes in the information flow between brain regions, involving in the episodic memory retrieval process. These convincing results emphasize that CREANN can be used as an appropriate method to estimate the causal relationship among brain signals.

The neural network classification of false killer whale (Pseudorca crassidens) vocalizations.

PubMed

Murray, S O; Mercado, E; Roitblat, H L

1998-12-01

This study reports the use of unsupervised, self-organizing neural network to categorize the repertoire of false killer whale vocalizations. Self-organizing networks are capable of detecting patterns in their input and partitioning those patterns into categories without requiring that the number or types of categories be predefined. The inputs for the neural networks were two-dimensional characterization of false killer whale vocalization, where each vocalization was characterized by a sequence of short-time measurements of duty cycle and peak frequency. The first neural network used competitive learning, where units in a competitive layer distributed themselves to recognize frequently presented input vectors. This network resulted in classes representing typical patterns in the vocalizations. The second network was a Kohonen feature map which organized the outputs topologically, providing a graphical organization of pattern relationships. The networks performed well as measured by (1) the average correlation between the input vectors and the weight vectors for each category, and (2) the ability of the networks to classify novel vocalizations. The techniques used in this study could easily be applied to other species and facilitate the development of objective, comprehensive repertoire models.

Fast, Simple and Accurate Handwritten Digit Classification by Training Shallow Neural Network Classifiers with the ‘Extreme Learning Machine’ Algorithm

PubMed Central

McDonnell, Mark D.; Tissera, Migel D.; Vladusich, Tony; van Schaik, André; Tapson, Jonathan

2015-01-01

Recent advances in training deep (multi-layer) architectures have inspired a renaissance in neural network use. For example, deep convolutional networks are becoming the default option for difficult tasks on large datasets, such as image and speech recognition. However, here we show that error rates below 1% on the MNIST handwritten digit benchmark can be replicated with shallow non-convolutional neural networks. This is achieved by training such networks using the ‘Extreme Learning Machine’ (ELM) approach, which also enables a very rapid training time (∼ 10 minutes). Adding distortions, as is common practise for MNIST, reduces error rates even further. Our methods are also shown to be capable of achieving less than 5.5% error rates on the NORB image database. To achieve these results, we introduce several enhancements to the standard ELM algorithm, which individually and in combination can significantly improve performance. The main innovation is to ensure each hidden-unit operates only on a randomly sized and positioned patch of each image. This form of random ‘receptive field’ sampling of the input ensures the input weight matrix is sparse, with about 90% of weights equal to zero. Furthermore, combining our methods with a small number of iterations of a single-batch backpropagation method can significantly reduce the number of hidden-units required to achieve a particular performance. Our close to state-of-the-art results for MNIST and NORB suggest that the ease of use and accuracy of the ELM algorithm for designing a single-hidden-layer neural network classifier should cause it to be given greater consideration either as a standalone method for simpler problems, or as the final classification stage in deep neural networks applied to more difficult problems. PMID:26262687

Modified-hybrid optical neural network filter for multiple object recognition within cluttered scenes

NASA Astrophysics Data System (ADS)

Kypraios, Ioannis; Young, Rupert C. D.; Chatwin, Chris R.

2009-08-01

Motivated by the non-linear interpolation and generalization abilities of the hybrid optical neural network filter between the reference and non-reference images of the true-class object we designed the modifiedhybrid optical neural network filter. We applied an optical mask to the hybrid optical neural network's filter input. The mask was built with the constant weight connections of a randomly chosen image included in the training set. The resulted design of the modified-hybrid optical neural network filter is optimized for performing best in cluttered scenes of the true-class object. Due to the shift invariance properties inherited by its correlator unit the filter can accommodate multiple objects of the same class to be detected within an input cluttered image. Additionally, the architecture of the neural network unit of the general hybrid optical neural network filter allows the recognition of multiple objects of different classes within the input cluttered image by modifying the output layer of the unit. We test the modified-hybrid optical neural network filter for multiple objects of the same and of different classes' recognition within cluttered input images and video sequences of cluttered scenes. The filter is shown to exhibit with a single pass over the input data simultaneously out-of-plane rotation, shift invariance and good clutter tolerance. It is able to successfully detect and classify correctly the true-class objects within background clutter for which there has been no previous training.

Neural Network-Based Sensor Validation for Turboshaft Engines

NASA Technical Reports Server (NTRS)

Moller, James C.; Litt, Jonathan S.; Guo, Ten-Huei

1998-01-01

Sensor failure detection, isolation, and accommodation using a neural network approach is described. An auto-associative neural network is configured to perform dimensionality reduction on the sensor measurement vector and provide estimated sensor values. The sensor validation scheme is applied in a simulation of the T700 turboshaft engine in closed loop operation. Performance is evaluated based on the ability to detect faults correctly and maintain stable and responsive engine operation. The set of sensor outputs used for engine control forms the network input vector. Analytical redundancy is verified by training networks of successively smaller bottleneck layer sizes. Training data generation and strategy are discussed. The engine maintained stable behavior in the presence of sensor hard failures. With proper selection of fault determination thresholds, stability was maintained in the presence of sensor soft failures.

Modular Neural Networks for Speech Recognition.

DTIC Science & Technology

1996-08-01

automatic speech rccogni- tion, understanding and translation since the early 1950’ s . Although researchers have demonstrated impressive results with...nodes. It serves only as a data source for the following hidden layer( s ). Finally, the networks output is computed by neurons in the output layer. The...following update rule for weights in the hidden layer: w (,,•+I) ("’) E/V S (W W k- = wj, -- 7 - / v It is easy to generalize the backpropagation

SYNAPTIC DEPRESSION IN DEEP NEURAL NETWORKS FOR SPEECH PROCESSING.

PubMed

Zhang, Wenhao; Li, Hanyu; Yang, Minda; Mesgarani, Nima

2016-03-01

A characteristic property of biological neurons is their ability to dynamically change the synaptic efficacy in response to variable input conditions. This mechanism, known as synaptic depression, significantly contributes to the formation of normalized representation of speech features. Synaptic depression also contributes to the robust performance of biological systems. In this paper, we describe how synaptic depression can be modeled and incorporated into deep neural network architectures to improve their generalization ability. We observed that when synaptic depression is added to the hidden layers of a neural network, it reduces the effect of changing background activity in the node activations. In addition, we show that when synaptic depression is included in a deep neural network trained for phoneme classification, the performance of the network improves under noisy conditions not included in the training phase. Our results suggest that more complete neuron models may further reduce the gap between the biological performance and artificial computing, resulting in networks that better generalize to novel signal conditions.

Multilayer neural networks with extensively many hidden units.

PubMed

Rosen-Zvi, M; Engel, A; Kanter, I

2001-08-13

The information processing abilities of a multilayer neural network with a number of hidden units scaling as the input dimension are studied using statistical mechanics methods. The mapping from the input layer to the hidden units is performed by general symmetric Boolean functions, whereas the hidden layer is connected to the output by either discrete or continuous couplings. Introducing an overlap in the space of Boolean functions as order parameter, the storage capacity is found to scale with the logarithm of the number of implementable Boolean functions. The generalization behavior is smooth for continuous couplings and shows a discontinuous transition to perfect generalization for discrete ones.

One-dimensional inversion of geo-electrical resistivity sounding data using artificial neural networks—a case study

NASA Astrophysics Data System (ADS)

Singh, U. K.; Tiwari, R. K.; Singh, S. B.

2005-02-01

This paper deals with the application of artificial neural networks (ANN) technique for the study of a case history using 1-D inversion of vertical electrical resistivity sounding (VES) data from the Puga valley, Kashmir, India. The study area is important for its rich geothermal resources as well as from the tectonic point of view as it is located near the collision boundary of the Indo-Asian crustal plates. In order to understand the resistivity structure and layer thicknesses, we used here three-layer feedforward neural networks to model and predict measured VES data. Three algorithms, e.g. back-propagation (BP), adaptive back-propagation (ABP) and Levenberg-Marquardt algorithm (LMA) were applied to the synthetic as well as real VES field data and efficiency of supervised training network are compared. Analyses suggest that LMA is computationally faster and give results, which are comparatively more accurate and consistent than BP and ABP. The results obtained using the ANN inversions are remarkably correlated with the available borehole litho-logs. The feasibility study suggests that ANN methods offer an excellent complementary tool for the direct detection of layered resistivity structure.

Feed-Forward Propagation of Temporal and Rate Information between Cortical Populations during Coherent Activation in Engineered In Vitro Networks.

PubMed

DeMarse, Thomas B; Pan, Liangbin; Alagapan, Sankaraleengam; Brewer, Gregory J; Wheeler, Bruce C

2016-01-01

Transient propagation of information across neuronal assembles is thought to underlie many cognitive processes. However, the nature of the neural code that is embedded within these transmissions remains uncertain. Much of our understanding of how information is transmitted among these assemblies has been derived from computational models. While these models have been instrumental in understanding these processes they often make simplifying assumptions about the biophysical properties of neurons that may influence the nature and properties expressed. To address this issue we created an in vitro analog of a feed-forward network composed of two small populations (also referred to as assemblies or layers) of living dissociated rat cortical neurons. The populations were separated by, and communicated through, a microelectromechanical systems (MEMS) device containing a strip of microscale tunnels. Delayed culturing of one population in the first layer followed by the second a few days later induced the unidirectional growth of axons through the microtunnels resulting in a primarily feed-forward communication between these two small neural populations. In this study we systematically manipulated the number of tunnels that connected each layer and hence, the number of axons providing communication between those populations. We then assess the effect of reducing the number of tunnels has upon the properties of between-layer communication capacity and fidelity of neural transmission among spike trains transmitted across and within layers. We show evidence based on Victor-Purpura's and van Rossum's spike train similarity metrics supporting the presence of both rate and temporal information embedded within these transmissions whose fidelity increased during communication both between and within layers when the number of tunnels are increased. We also provide evidence reinforcing the role of synchronized activity upon transmission fidelity during the spontaneous synchronized network burst events that propagated between layers and highlight the potential applications of these MEMs devices as a tool for further investigation of structure and functional dynamics among neural populations.

Feed-Forward Propagation of Temporal and Rate Information between Cortical Populations during Coherent Activation in Engineered In Vitro Networks

PubMed Central

DeMarse, Thomas B.; Pan, Liangbin; Alagapan, Sankaraleengam; Brewer, Gregory J.; Wheeler, Bruce C.

2016-01-01

Transient propagation of information across neuronal assembles is thought to underlie many cognitive processes. However, the nature of the neural code that is embedded within these transmissions remains uncertain. Much of our understanding of how information is transmitted among these assemblies has been derived from computational models. While these models have been instrumental in understanding these processes they often make simplifying assumptions about the biophysical properties of neurons that may influence the nature and properties expressed. To address this issue we created an in vitro analog of a feed-forward network composed of two small populations (also referred to as assemblies or layers) of living dissociated rat cortical neurons. The populations were separated by, and communicated through, a microelectromechanical systems (MEMS) device containing a strip of microscale tunnels. Delayed culturing of one population in the first layer followed by the second a few days later induced the unidirectional growth of axons through the microtunnels resulting in a primarily feed-forward communication between these two small neural populations. In this study we systematically manipulated the number of tunnels that connected each layer and hence, the number of axons providing communication between those populations. We then assess the effect of reducing the number of tunnels has upon the properties of between-layer communication capacity and fidelity of neural transmission among spike trains transmitted across and within layers. We show evidence based on Victor-Purpura’s and van Rossum’s spike train similarity metrics supporting the presence of both rate and temporal information embedded within these transmissions whose fidelity increased during communication both between and within layers when the number of tunnels are increased. We also provide evidence reinforcing the role of synchronized activity upon transmission fidelity during the spontaneous synchronized network burst events that propagated between layers and highlight the potential applications of these MEMs devices as a tool for further investigation of structure and functional dynamics among neural populations. PMID:27147977

Wishart Deep Stacking Network for Fast POLSAR Image Classification.

PubMed

Jiao, Licheng; Liu, Fang

2016-05-11

Inspired by the popular deep learning architecture - Deep Stacking Network (DSN), a specific deep model for polarimetric synthetic aperture radar (POLSAR) image classification is proposed in this paper, which is named as Wishart Deep Stacking Network (W-DSN). First of all, a fast implementation of Wishart distance is achieved by a special linear transformation, which speeds up the classification of POLSAR image and makes it possible to use this polarimetric information in the following Neural Network (NN). Then a single-hidden-layer neural network based on the fast Wishart distance is defined for POLSAR image classification, which is named as Wishart Network (WN) and improves the classification accuracy. Finally, a multi-layer neural network is formed by stacking WNs, which is in fact the proposed deep learning architecture W-DSN for POLSAR image classification and improves the classification accuracy further. In addition, the structure of WN can be expanded in a straightforward way by adding hidden units if necessary, as well as the structure of the W-DSN. As a preliminary exploration on formulating specific deep learning architecture for POLSAR image classification, the proposed methods may establish a simple but clever connection between POLSAR image interpretation and deep learning. The experiment results tested on real POLSAR image show that the fast implementation of Wishart distance is very efficient (a POLSAR image with 768000 pixels can be classified in 0.53s), and both the single-hidden-layer architecture WN and the deep learning architecture W-DSN for POLSAR image classification perform well and work efficiently.

A Software Package for Neural Network Applications Development

NASA Technical Reports Server (NTRS)

Baran, Robert H.

1993-01-01

Original Backprop (Version 1.2) is an MS-DOS package of four stand-alone C-language programs that enable users to develop neural network solutions to a variety of practical problems. Original Backprop generates three-layer, feed-forward (series-coupled) networks which map fixed-length input vectors into fixed length output vectors through an intermediate (hidden) layer of binary threshold units. Version 1.2 can handle up to 200 input vectors at a time, each having up to 128 real-valued components. The first subprogram, TSET, appends a number (up to 16) of classification bits to each input, thus creating a training set of input output pairs. The second subprogram, BACKPROP, creates a trilayer network to do the prescribed mapping and modifies the weights of its connections incrementally until the training set is leaned. The learning algorithm is the 'back-propagating error correction procedures first described by F. Rosenblatt in 1961. The third subprogram, VIEWNET, lets the trained network be examined, tested, and 'pruned' (by the deletion of unnecessary hidden units). The fourth subprogram, DONET, makes a TSR routine by which the finished product of the neural net design-and-training exercise can be consulted under other MS-DOS applications.

Comparing success levels of different neural network structures in extracting discriminative information from the response patterns of a temperature-modulated resistive gas sensor

NASA Astrophysics Data System (ADS)

Hosseini-Golgoo, S. M.; Bozorgi, H.; Saberkari, A.

2015-06-01

Performances of three neural networks, consisting of a multi-layer perceptron, a radial basis function, and a neuro-fuzzy network with local linear model tree training algorithm, in modeling and extracting discriminative features from the response patterns of a temperature-modulated resistive gas sensor are quantitatively compared. For response pattern recording, a voltage staircase containing five steps each with a 20 s plateau is applied to the micro-heater of the sensor, when 12 different target gases, each at 11 concentration levels, are present. In each test, the hidden layer neuron weights are taken as the discriminatory feature vector of the target gas. These vectors are then mapped to a 3D feature space using linear discriminant analysis. The discriminative information content of the feature vectors are determined by the calculation of the Fisher’s discriminant ratio, affording quantitative comparison among the success rates achieved by the different neural network structures. The results demonstrate a superior discrimination ratio for features extracted from local linear neuro-fuzzy and radial-basis-function networks with recognition rates of 96.27% and 90.74%, respectively.

Noise-tolerant inverse analysis models for nondestructive evaluation of transportation infrastructure systems using neural networks

NASA Astrophysics Data System (ADS)

Ceylan, Halil; Gopalakrishnan, Kasthurirangan; Birkan Bayrak, Mustafa; Guclu, Alper

2013-09-01

The need to rapidly and cost-effectively evaluate the present condition of pavement infrastructure is a critical issue concerning the deterioration of ageing transportation infrastructure all around the world. Nondestructive testing (NDT) and evaluation methods are well-suited for characterising materials and determining structural integrity of pavement systems. The falling weight deflectometer (FWD) is a NDT equipment used to assess the structural condition of highway and airfield pavement systems and to determine the moduli of pavement layers. This involves static or dynamic inverse analysis (referred to as backcalculation) of FWD deflection profiles in the pavement surface under a simulated truck load. The main objective of this study was to employ biologically inspired computational systems to develop robust pavement layer moduli backcalculation algorithms that can tolerate noise or inaccuracies in the FWD deflection data collected in the field. Artificial neural systems, also known as artificial neural networks (ANNs), are valuable computational intelligence tools that are increasingly being used to solve resource-intensive complex engineering problems. Unlike the linear elastic layered theory commonly used in pavement layer backcalculation, non-linear unbound aggregate base and subgrade soil response models were used in an axisymmetric finite element structural analysis programme to generate synthetic database for training and testing the ANN models. In order to develop more robust networks that can tolerate the noisy or inaccurate pavement deflection patterns in the NDT data, several network architectures were trained with varying levels of noise in them. The trained ANN models were capable of rapidly predicting the pavement layer moduli and critical pavement responses (tensile strains at the bottom of the asphalt concrete layer, compressive strains on top of the subgrade layer and the deviator stresses on top of the subgrade layer), and also pavement surface deflections with very low average errors comparable with those obtained directly from the finite element analyses.

Hermite Functional Link Neural Network for Solving the Van der Pol-Duffing Oscillator Equation.

PubMed

Mall, Susmita; Chakraverty, S

2016-08-01

Hermite polynomial-based functional link artificial neural network (FLANN) is proposed here to solve the Van der Pol-Duffing oscillator equation. A single-layer hermite neural network (HeNN) model is used, where a hidden layer is replaced by expansion block of input pattern using Hermite orthogonal polynomials. A feedforward neural network model with the unsupervised error backpropagation principle is used for modifying the network parameters and minimizing the computed error function. The Van der Pol-Duffing and Duffing oscillator equations may not be solved exactly. Here, approximate solutions of these types of equations have been obtained by applying the HeNN model for the first time. Three mathematical example problems and two real-life application problems of Van der Pol-Duffing oscillator equation, extracting the features of early mechanical failure signal and weak signal detection problems, are solved using the proposed HeNN method. HeNN approximate solutions have been compared with results obtained by the well known Runge-Kutta method. Computed results are depicted in term of graphs. After training the HeNN model, we may use it as a black box to get numerical results at any arbitrary point in the domain. Thus, the proposed HeNN method is efficient. The results reveal that this method is reliable and can be applied to other nonlinear problems too.

Completing sparse and disconnected protein-protein network by deep learning.

PubMed

Huang, Lei; Liao, Li; Wu, Cathy H

2018-03-22

Protein-protein interaction (PPI) prediction remains a central task in systems biology to achieve a better and holistic understanding of cellular and intracellular processes. Recently, an increasing number of computational methods have shifted from pair-wise prediction to network level prediction. Many of the existing network level methods predict PPIs under the assumption that the training network should be connected. However, this assumption greatly affects the prediction power and limits the application area because the current golden standard PPI networks are usually very sparse and disconnected. Therefore, how to effectively predict PPIs based on a training network that is sparse and disconnected remains a challenge. In this work, we developed a novel PPI prediction method based on deep learning neural network and regularized Laplacian kernel. We use a neural network with an autoencoder-like architecture to implicitly simulate the evolutionary processes of a PPI network. Neurons of the output layer correspond to proteins and are labeled with values (1 for interaction and 0 for otherwise) from the adjacency matrix of a sparse disconnected training PPI network. Unlike autoencoder, neurons at the input layer are given all zero input, reflecting an assumption of no a priori knowledge about PPIs, and hidden layers of smaller sizes mimic ancient interactome at different times during evolution. After the training step, an evolved PPI network whose rows are outputs of the neural network can be obtained. We then predict PPIs by applying the regularized Laplacian kernel to the transition matrix that is built upon the evolved PPI network. The results from cross-validation experiments show that the PPI prediction accuracies for yeast data and human data measured as AUC are increased by up to 8.4 and 14.9% respectively, as compared to the baseline. Moreover, the evolved PPI network can also help us leverage complementary information from the disconnected training network and multiple heterogeneous data sources. Tested by the yeast data with six heterogeneous feature kernels, the results show our method can further improve the prediction performance by up to 2%, which is very close to an upper bound that is obtained by an Approximate Bayesian Computation based sampling method. The proposed evolution deep neural network, coupled with regularized Laplacian kernel, is an effective tool in completing sparse and disconnected PPI networks and in facilitating integration of heterogeneous data sources.

Training Deep Convolutional Neural Networks with Resistive Cross-Point Devices

PubMed Central

Gokmen, Tayfun; Onen, Murat; Haensch, Wilfried

2017-01-01

In a previous work we have detailed the requirements for obtaining maximal deep learning performance benefit by implementing fully connected deep neural networks (DNN) in the form of arrays of resistive devices. Here we extend the concept of Resistive Processing Unit (RPU) devices to convolutional neural networks (CNNs). We show how to map the convolutional layers to fully connected RPU arrays such that the parallelism of the hardware can be fully utilized in all three cycles of the backpropagation algorithm. We find that the noise and bound limitations imposed by the analog nature of the computations performed on the arrays significantly affect the training accuracy of the CNNs. Noise and bound management techniques are presented that mitigate these problems without introducing any additional complexity in the analog circuits and that can be addressed by the digital circuits. In addition, we discuss digitally programmable update management and device variability reduction techniques that can be used selectively for some of the layers in a CNN. We show that a combination of all those techniques enables a successful application of the RPU concept for training CNNs. The techniques discussed here are more general and can be applied beyond CNN architectures and therefore enables applicability of the RPU approach to a large class of neural network architectures. PMID:29066942

Training Deep Convolutional Neural Networks with Resistive Cross-Point Devices.

PubMed

Gokmen, Tayfun; Onen, Murat; Haensch, Wilfried

2017-01-01

In a previous work we have detailed the requirements for obtaining maximal deep learning performance benefit by implementing fully connected deep neural networks (DNN) in the form of arrays of resistive devices. Here we extend the concept of Resistive Processing Unit (RPU) devices to convolutional neural networks (CNNs). We show how to map the convolutional layers to fully connected RPU arrays such that the parallelism of the hardware can be fully utilized in all three cycles of the backpropagation algorithm. We find that the noise and bound limitations imposed by the analog nature of the computations performed on the arrays significantly affect the training accuracy of the CNNs. Noise and bound management techniques are presented that mitigate these problems without introducing any additional complexity in the analog circuits and that can be addressed by the digital circuits. In addition, we discuss digitally programmable update management and device variability reduction techniques that can be used selectively for some of the layers in a CNN. We show that a combination of all those techniques enables a successful application of the RPU concept for training CNNs. The techniques discussed here are more general and can be applied beyond CNN architectures and therefore enables applicability of the RPU approach to a large class of neural network architectures.

Near real-time analysis of extrinsic Fabry-Perot interferometric sensors under damped vibration using artificial neural networks

NASA Astrophysics Data System (ADS)

Dua, Rohit; Watkins, Steve E.

2009-03-01

Strain analysis due to vibration can provide insight into structural health. An Extrinsic Fabry-Perot Interferometric (EFPI) sensor under vibrational strain generates a non-linear modulated output. Advanced signal processing techniques, to extract important information such as absolute strain, are required to demodulate this non-linear output. Past research has employed Artificial Neural Networks (ANN) and Fast Fourier Transforms (FFT) to demodulate the EFPI sensor for limited conditions. These demodulation systems could only handle variations in absolute value of strain and frequency of actuation during a vibration event. This project uses an ANN approach to extend the demodulation system to include the variation in the damping coefficient of the actuating vibration, in a near real-time vibration scenario. A computer simulation provides training and testing data for the theoretical output of the EFPI sensor to demonstrate the approaches. FFT needed to be performed on a window of the EFPI output data. A small window of observation is obtained, while maintaining low absolute-strain prediction errors, heuristically. Results are obtained and compared from employing different ANN architectures including multi-layered feedforward ANN trained using Backpropagation Neural Network (BPNN), and Generalized Regression Neural Networks (GRNN). A two-layered algorithm fusion system is developed and tested that yields better results.

Analysis and Modeling of DIII-D Experiments With OMFIT and Neural Networks

NASA Astrophysics Data System (ADS)

Meneghini, O.; Luna, C.; Smith, S. P.; Lao, L. L.; GA Theory Team

2013-10-01

The OMFIT integrated modeling framework is designed to facilitate experimental data analysis and enable integrated simulations. This talk introduces this framework and presents a selection of its applications to the DIII-D experiment. Examples include kinetic equilibrium reconstruction analysis; evaluation of MHD stability in the core and in the edge; and self-consistent predictive steady-state transport modeling. The OMFIT framework also provides the platform for an innovative approach based on neural networks to predict electron and ion energy fluxes. In our study a multi-layer feed-forward back-propagation neural network is built and trained over a database of DIII-D data. It is found that given the same parameters that the highest fidelity models use, the neural network model is able to predict to a large degree the heat transport profiles observed in the DIII-D experiments. Once the network is built, the numerical cost of evaluating the transport coefficients is virtually nonexistent, thus making the neural network model particularly well suited for plasma control and quick exploration of operational scenarios. The implementation of the neural network model and benchmark with experimental results and gyro-kinetic models will be discussed. Work supported in part by the US DOE under DE-FG02-95ER54309.

Multi-modality image fusion based on enhanced fuzzy radial basis function neural networks.

PubMed

Chao, Zhen; Kim, Dohyeon; Kim, Hee-Joung

2018-04-01

In clinical applications, single modality images do not provide sufficient diagnostic information. Therefore, it is necessary to combine the advantages or complementarities of different modalities of images. Recently, neural network technique was applied to medical image fusion by many researchers, but there are still many deficiencies. In this study, we propose a novel fusion method to combine multi-modality medical images based on the enhanced fuzzy radial basis function neural network (Fuzzy-RBFNN), which includes five layers: input, fuzzy partition, front combination, inference, and output. Moreover, we propose a hybrid of the gravitational search algorithm (GSA) and error back propagation algorithm (EBPA) to train the network to update the parameters of the network. Two different patterns of images are used as inputs of the neural network, and the output is the fused image. A comparison with the conventional fusion methods and another neural network method through subjective observation and objective evaluation indexes reveals that the proposed method effectively synthesized the information of input images and achieved better results. Meanwhile, we also trained the network by using the EBPA and GSA, individually. The results reveal that the EBPGSA not only outperformed both EBPA and GSA, but also trained the neural network more accurately by analyzing the same evaluation indexes. Copyright © 2018 Associazione Italiana di Fisica Medica. Published by Elsevier Ltd. All rights reserved.

Efficient airport detection using region-based fully convolutional neural networks

NASA Astrophysics Data System (ADS)

Xin, Peng; Xu, Yuelei; Zhang, Xulei; Ma, Shiping; Li, Shuai; Lv, Chao

2018-04-01

This paper presents a model for airport detection using region-based fully convolutional neural networks. To achieve fast detection with high accuracy, we shared the conv layers between the region proposal procedure and the airport detection procedure and used graphics processing units (GPUs) to speed up the training and testing time. For lack of labeled data, we transferred the convolutional layers of ZF net pretrained by ImageNet to initialize the shared convolutional layers, then we retrained the model using the alternating optimization training strategy. The proposed model has been tested on an airport dataset consisting of 600 images. Experiments show that the proposed method can distinguish airports in our dataset from similar background scenes almost real-time with high accuracy, which is much better than traditional methods.

Forecasting Flare Activity Using Deep Convolutional Neural Networks

NASA Astrophysics Data System (ADS)

Hernandez, T.

2017-12-01

Current operational flare forecasting relies on human morphological analysis of active regions and the persistence of solar flare activity through time (i.e. that the Sun will continue to do what it is doing right now: flaring or remaining calm). In this talk we present the results of applying deep Convolutional Neural Networks (CNNs) to the problem of solar flare forecasting. CNNs operate by training a set of tunable spatial filters that, in combination with neural layer interconnectivity, allow CNNs to automatically identify significant spatial structures predictive for classification and regression problems. We will start by discussing the applicability and success rate of the approach, the advantages it has over non-automated forecasts, and how mining our trained neural network provides a fresh look into the mechanisms behind magnetic energy storage and release.

Neuro-classification of multi-type Landsat Thematic Mapper data

NASA Technical Reports Server (NTRS)

Zhuang, Xin; Engel, Bernard A.; Fernandez, R. N.; Johannsen, Chris J.

1991-01-01

Neural networks have been successful in image classification and have shown potential for classifying remotely sensed data. This paper presents classifications of multitype Landsat Thematic Mapper (TM) data using neural networks. The Landsat TM Image for March 23, 1987 with accompanying ground observation data for a study area In Miami County, Indiana, U.S.A. was utilized to assess recognition of crop residues. Principal components and spectral ratio transformations were performed on the TM data. In addition, a layer of the geographic information system (GIS) for the study site was incorporated to generate GIS-enhanced TM data. This paper discusses (1) the performance of neuro-classification on each type of data, (2) how neural networks recognized each type of data as a new image and (3) comparisons of the results for each type of data obtained using neural networks, maximum likelihood, and minimum distance classifiers.

Seismic activity prediction using computational intelligence techniques in northern Pakistan

NASA Astrophysics Data System (ADS)

Asim, Khawaja M.; Awais, Muhammad; Martínez-Álvarez, F.; Iqbal, Talat

2017-10-01

Earthquake prediction study is carried out for the region of northern Pakistan. The prediction methodology includes interdisciplinary interaction of seismology and computational intelligence. Eight seismic parameters are computed based upon the past earthquakes. Predictive ability of these eight seismic parameters is evaluated in terms of information gain, which leads to the selection of six parameters to be used in prediction. Multiple computationally intelligent models have been developed for earthquake prediction using selected seismic parameters. These models include feed-forward neural network, recurrent neural network, random forest, multi layer perceptron, radial basis neural network, and support vector machine. The performance of every prediction model is evaluated and McNemar's statistical test is applied to observe the statistical significance of computational methodologies. Feed-forward neural network shows statistically significant predictions along with accuracy of 75% and positive predictive value of 78% in context of northern Pakistan.

Development of a neural network for early detection of renal osteodystrophy

NASA Astrophysics Data System (ADS)

Cheng, Shirley N.; Chan, Heang-Ping; Adler, Ronald; Niklason, Loren T.; Chang, Chair-Li

1991-07-01

Bone erosion presenting as subperiosteal resorption on the phalanges of the hand is an early manifestation of hyperparathyroidism associated with chronic renal failure. At present, the diagnosis is made by trained radiologists through visual inspection of hand radiographs. In this study, a neural network is being developed to assess the feasibility of computer-aided detection of these changes. A two-pass approach is adopted. The digitized image is first compressed by a Laplacian pyramid compact code. The first neural network locates the region of interest using vertical projections along the phalanges and then the horizontal projections across the phalanges. A second neural network is used to classify texture variations of trabecular patterns in the region using a concurrence matrix as the input to a two-dimensional sensor layer to detect the degree of associated osteopenia. Preliminary results demonstrate the feasibility of this approach.

Identifying Jets Using Artifical Neural Networks

NASA Astrophysics Data System (ADS)

Rosand, Benjamin; Caines, Helen; Checa, Sofia

2017-09-01

We investigate particle jet interactions with the Quark Gluon Plasma (QGP) using artificial neural networks modeled on those used in computer image recognition. We create jet images by binning jet particles into pixels and preprocessing every image. We analyzed the jets with a Multi-layered maxout network and a convolutional network. We demonstrate each network's effectiveness in differentiating simulated quenched jets from unquenched jets, and we investigate the method that the network uses to discriminate among different quenched jet simulations. Finally, we develop a greater understanding of the physics behind quenched jets by investigating what the network learnt as well as its effectiveness in differentiating samples. Yale College Freshman Summer Research Fellowship in the Sciences and Engineering.

Autonomous self-configuration of artificial neural networks for data classification or system control

NASA Astrophysics Data System (ADS)

Fink, Wolfgang

2009-05-01

Artificial neural networks (ANNs) are powerful methods for the classification of multi-dimensional data as well as for the control of dynamic systems. In general terms, ANNs consist of neurons that are, e.g., arranged in layers and interconnected by real-valued or binary neural couplings or weights. ANNs try mimicking the processing taking place in biological brains. The classification and generalization capabilities of ANNs are given by the interconnection architecture and the coupling strengths. To perform a certain classification or control task with a particular ANN architecture (i.e., number of neurons, number of layers, etc.), the inter-neuron couplings and their accordant coupling strengths must be determined (1) either by a priori design (i.e., manually) or (2) using training algorithms such as error back-propagation. The more complex the classification or control task, the less obvious it is how to determine an a priori design of an ANN, and, as a consequence, the architecture choice becomes somewhat arbitrary. Furthermore, rather than being able to determine for a given architecture directly the corresponding coupling strengths necessary to perform the classification or control task, these have to be obtained/learned through training of the ANN on test data. We report on the use of a Stochastic Optimization Framework (SOF; Fink, SPIE 2008) for the autonomous self-configuration of Artificial Neural Networks (i.e., the determination of number of hidden layers, number of neurons per hidden layer, interconnections between neurons, and respective coupling strengths) for performing classification or control tasks. This may provide an approach towards cognizant and self-adapting computing architectures and systems.

Learning, memory, and the role of neural network architecture.

PubMed

Hermundstad, Ann M; Brown, Kevin S; Bassett, Danielle S; Carlson, Jean M

2011-06-01

The performance of information processing systems, from artificial neural networks to natural neuronal ensembles, depends heavily on the underlying system architecture. In this study, we compare the performance of parallel and layered network architectures during sequential tasks that require both acquisition and retention of information, thereby identifying tradeoffs between learning and memory processes. During the task of supervised, sequential function approximation, networks produce and adapt representations of external information. Performance is evaluated by statistically analyzing the error in these representations while varying the initial network state, the structure of the external information, and the time given to learn the information. We link performance to complexity in network architecture by characterizing local error landscape curvature. We find that variations in error landscape structure give rise to tradeoffs in performance; these include the ability of the network to maximize accuracy versus minimize inaccuracy and produce specific versus generalizable representations of information. Parallel networks generate smooth error landscapes with deep, narrow minima, enabling them to find highly specific representations given sufficient time. While accurate, however, these representations are difficult to generalize. In contrast, layered networks generate rough error landscapes with a variety of local minima, allowing them to quickly find coarse representations. Although less accurate, these representations are easily adaptable. The presence of measurable performance tradeoffs in both layered and parallel networks has implications for understanding the behavior of a wide variety of natural and artificial learning systems.

Prediction of forced expiratory volume in pulmonary function test using radial basis neural networks and k-means clustering.

PubMed

Manoharan, Sujatha C; Ramakrishnan, Swaminathan

2009-10-01

In this work, prediction of forced expiratory volume in pulmonary function test, carried out using spirometry and neural networks is presented. The pulmonary function data were recorded from volunteers using commercial available flow volume spirometer in standard acquisition protocol. The Radial Basis Function neural networks were used to predict forced expiratory volume in 1 s (FEV1) from the recorded flow volume curves. The optimal centres of the hidden layer of radial basis function were determined by k-means clustering algorithm. The performance of the neural network model was evaluated by computing their prediction error statistics of average value, standard deviation, root mean square and their correlation with the true data for normal, restrictive and obstructive cases. Results show that the adopted neural networks are capable of predicting FEV1 in both normal and abnormal cases. Prediction accuracy was more in obstructive abnormality when compared to restrictive cases. It appears that this method of assessment is useful in diagnosing the pulmonary abnormalities with incomplete data and data with poor recording.

Real-time camera-based face detection using a modified LAMSTAR neural network system

NASA Astrophysics Data System (ADS)

Girado, Javier I.; Sandin, Daniel J.; DeFanti, Thomas A.; Wolf, Laura K.

2003-03-01

This paper describes a cost-effective, real-time (640x480 at 30Hz) upright frontal face detector as part of an ongoing project to develop a video-based, tetherless 3D head position and orientation tracking system. The work is specifically targeted for auto-stereoscopic displays and projection-based virtual reality systems. The proposed face detector is based on a modified LAMSTAR neural network system. At the input stage, after achieving image normalization and equalization, a sub-window analyzes facial features using a neural network. The sub-window is segmented, and each part is fed to a neural network layer consisting of a Kohonen Self-Organizing Map (SOM). The output of the SOM neural networks are interconnected and related by correlation-links, and can hence determine the presence of a face with enough redundancy to provide a high detection rate. To avoid tracking multiple faces simultaneously, the system is initially trained to track only the face centered in a box superimposed on the display. The system is also rotationally and size invariant to a certain degree.

Logarithmic r-θ mapping for hybrid optical neural network filter for multiple objects recognition within cluttered scenes

NASA Astrophysics Data System (ADS)

Kypraios, Ioannis; Young, Rupert C. D.; Chatwin, Chris R.; Birch, Phil M.

2009-04-01

θThe window unit in the design of the complex logarithmic r-θ mapping for hybrid optical neural network filter can allow multiple objects of the same class to be detected within the input image. Additionally, the architecture of the neural network unit of the complex logarithmic r-θ mapping for hybrid optical neural network filter becomes attractive for accommodating the recognition of multiple objects of different classes within the input image by modifying the output layer of the unit. We test the overall filter for multiple objects of the same and of different classes' recognition within cluttered input images and video sequences of cluttered scenes. Logarithmic r-θ mapping for hybrid optical neural network filter is shown to exhibit with a single pass over the input data simultaneously in-plane rotation, out-of-plane rotation, scale, log r-θ map translation and shift invariance, and good clutter tolerance by recognizing correctly the different objects within the cluttered scenes. We record in our results additional extracted information from the cluttered scenes about the objects' relative position, scale and in-plane rotation.

Hacia la predicción del Número R de Wolf de manchas solares utilizando Redes Neuronales con retardos temporales

NASA Astrophysics Data System (ADS)

Francile, C.; Luoni, M. L.

We present a prediction of the time series of the Wolf number R of sunspots using "time lagged feed forward neural networks". We use two types of networks: the focused and distributed ones which were trained with the back propagation of errors algorithm and the temporal back propagation algorithm respectively. As inputs to neural networks we use the time series of the number R averaged annually and monthly with the method IR5. As data sets for training and test we choose certain intervals of the time series similar to other works, in order to compare the results. Finally we discuss the topology of the networks used, the number of delays used, the number of neurons per layer, the number of hidden layers and the results in the prediction of the series between one and six steps ahead. FULL TEXT IN SPANISH

Neural network and letter recognition

DOE Office of Scientific and Technical Information (OSTI.GOV)

Lee, Hue Yeon.

Neural net architectures and learning algorithms that recognize hand written 36 alphanumeric characters are studied. The thin line input patterns written in 32 x 32 binary array are used. The system is comprised of two major components, viz. a preprocessing unit and a Recognition unit. The preprocessing unit in turn consists of three layers of neurons; the U-layer, the V-layer, and the C-layer. The functions of the U-layer is to extract local features by template matching. The correlation between the detected local features are considered. Through correlating neurons in a plane with their neighboring neurons, the V-layer would thicken themore » on-cells or lines that are groups of on-cells of the previous layer. These two correlations would yield some deformation tolerance and some of the rotational tolerance of the system. The C-layer then compresses data through the Gabor transform. Pattern dependent choice of center and wavelengths of Gabor filters is the cause of shift and scale tolerance of the system. Three different learning schemes had been investigated in the recognition unit, namely; the error back propagation learning with hidden units, a simple perceptron learning, and a competitive learning. Their performances were analyzed and compared. Since sometimes the network fails to distinguish between two letters that are inherently similar, additional ambiguity resolving neural nets are introduced on top of the above main neural net. The two dimensional Fourier transform is used as the preprocessing and the perceptron is used as the recognition unit of the ambiguity resolver. One hundred different person's handwriting sets are collected. Some of these are used as the training sets and the remainders are used as the test sets.« less

Dependency-based long short term memory network for drug-drug interaction extraction.

PubMed

Wang, Wei; Yang, Xi; Yang, Canqun; Guo, Xiaowei; Zhang, Xiang; Wu, Chengkun

2017-12-28

Drug-drug interaction extraction (DDI) needs assistance from automated methods to address the explosively increasing biomedical texts. In recent years, deep neural network based models have been developed to address such needs and they have made significant progress in relation identification. We propose a dependency-based deep neural network model for DDI extraction. By introducing the dependency-based technique to a bi-directional long short term memory network (Bi-LSTM), we build three channels, namely, Linear channel, DFS channel and BFS channel. All of these channels are constructed with three network layers, including embedding layer, LSTM layer and max pooling layer from bottom up. In the embedding layer, we extract two types of features, one is distance-based feature and another is dependency-based feature. In the LSTM layer, a Bi-LSTM is instituted in each channel to better capture relation information. Then max pooling is used to get optimal features from the entire encoding sequential data. At last, we concatenate the outputs of all channels and then link it to the softmax layer for relation identification. To the best of our knowledge, our model achieves new state-of-the-art performance with the F-score of 72.0% on the DDIExtraction 2013 corpus. Moreover, our approach obtains much higher Recall value compared to the existing methods. The dependency-based Bi-LSTM model can learn effective relation information with less feature engineering in the task of DDI extraction. Besides, the experimental results show that our model excels at balancing the Precision and Recall values.

An architecture for designing fuzzy logic controllers using neural networks

NASA Technical Reports Server (NTRS)

Berenji, Hamid R.

1991-01-01

Described here is an architecture for designing fuzzy controllers through a hierarchical process of control rule acquisition and by using special classes of neural network learning techniques. A new method for learning to refine a fuzzy logic controller is introduced. A reinforcement learning technique is used in conjunction with a multi-layer neural network model of a fuzzy controller. The model learns by updating its prediction of the plant's behavior and is related to the Sutton's Temporal Difference (TD) method. The method proposed here has the advantage of using the control knowledge of an experienced operator and fine-tuning it through the process of learning. The approach is applied to a cart-pole balancing system.

Neural network diagnosis of avascular necrosis from magnetic resonance images

NASA Astrophysics Data System (ADS)

Manduca, Armando; Christy, Paul S.; Ehman, Richard L.

1993-09-01

We have explored the use of artificial neural networks to diagnose avascular necrosis (AVN) of the femoral head from magnetic resonance images. We have developed multi-layer perceptron networks, trained with conjugate gradient optimization, which diagnose AVN from single sagittal images of the femoral head with 100% accuracy on the training data and 97% accuracy on test data. These networks use only the raw image as input (with minimal preprocessing to average the images down to 32 X 32 size and to scale the input data values) and learn to extract their own features for the diagnosis decision. Various experiments with these networks are described.

Analysis of Artificial Neural Network in Erosion Modeling: A Case Study of Serang Watershed

NASA Astrophysics Data System (ADS)

Arif, N.; Danoedoro, P.; Hartono

2017-12-01

Erosion modeling is an important measuring tool for both land users and decision makers to evaluate land cultivation and thus it is necessary to have a model to represent the actual reality. Erosion models are a complex model because of uncertainty data with different sources and processing procedures. Artificial neural networks can be relied on for complex and non-linear data processing such as erosion data. The main difficulty in artificial neural network training is the determination of the value of each network input parameters, i.e. hidden layer, momentum, learning rate, momentum, and RMS. This study tested the capability of artificial neural network application in the prediction of erosion risk with some input parameters through multiple simulations to get good classification results. The model was implemented in Serang Watershed, Kulonprogo, Yogyakarta which is one of the critical potential watersheds in Indonesia. The simulation results showed the number of iterations that gave a significant effect on the accuracy compared to other parameters. A small number of iterations can produce good accuracy if the combination of other parameters was right. In this case, one hidden layer was sufficient to produce good accuracy. The highest training accuracy achieved in this study was 99.32%, occurred in ANN 14 simulation with combination of network input parameters of 1 HL; LR 0.01; M 0.5; RMS 0.0001, and the number of iterations of 15000. The ANN training accuracy was not influenced by the number of channels, namely input dataset (erosion factors) as well as data dimensions, rather it was determined by changes in network parameters.

Alcoholism Detection by Data Augmentation and Convolutional Neural Network with Stochastic Pooling.

PubMed

Wang, Shui-Hua; Lv, Yi-Ding; Sui, Yuxiu; Liu, Shuai; Wang, Su-Jing; Zhang, Yu-Dong

2017-11-17

Alcohol use disorder (AUD) is an important brain disease. It alters the brain structure. Recently, scholars tend to use computer vision based techniques to detect AUD. We collected 235 subjects, 114 alcoholic and 121 non-alcoholic. Among the 235 image, 100 images were used as training set, and data augmentation method was used. The rest 135 images were used as test set. Further, we chose the latest powerful technique-convolutional neural network (CNN) based on convolutional layer, rectified linear unit layer, pooling layer, fully connected layer, and softmax layer. We also compared three different pooling techniques: max pooling, average pooling, and stochastic pooling. The results showed that our method achieved a sensitivity of 96.88%, a specificity of 97.18%, and an accuracy of 97.04%. Our method was better than three state-of-the-art approaches. Besides, stochastic pooling performed better than other max pooling and average pooling. We validated CNN with five convolution layers and two fully connected layers performed the best. The GPU yielded a 149× acceleration in training and a 166× acceleration in test, compared to CPU.

Inversion of quasi-3D DC resistivity imaging data using artificial neural networks

NASA Astrophysics Data System (ADS)

Neyamadpour, Ahmad; Wan Abdullah, W. A. T.; Taib, Samsudin

2010-02-01

The objective of this paper is to investigate the applicability of artificial neural networks in inverting quasi-3D DC resistivity imaging data. An electrical resistivity imaging survey was carried out along seven parallel lines using a dipole-dipole array to confirm the validation of the results of an inversion using an artificial neural network technique. The model used to produce synthetic data to train the artificial neural network was a homogeneous medium of 100Ωm resistivity with an embedded anomalous body of 1000Ωm resistivity. The network was trained using 21 datasets (comprising 12159 data points) and tested on another 11 synthetic datasets (comprising 6369 data points) and on real field data. Another 24 test datasets (comprising 13896 data points) consisting of different resistivities for the background and the anomalous bodies were used in order to test the interpolation and extrapolation of network properties. Different learning paradigms were tried in the training process of the neural network, with the resilient propagation paradigm being the most efficient. The number of nodes, hidden layers, and efficient values for learning rate and momentum coefficient have been studied. Although a significant correlation between results of the neural network and the conventional robust inversion technique was found, the ANN results show more details of the subsurface structure, and the RMS misfits for the results of the neural network are less than seen with conventional methods. The interpreted results show that the trained network was able to invert quasi-3D electrical resistivity imaging data obtained by dipole-dipole configuration both rapidly and accurately.

Beyond Fine Tuning: Adding capacity to leverage few labels

DOE Office of Scientific and Technical Information (OSTI.GOV)

Hodas, Nathan O.; Shaffer, Kyle J.; Yankov, Artem

2017-12-09

In this paper we present a technique to train neural network models on small amounts of data. Current methods for training neural networks on small amounts of rich data typically rely on strategies such as fine-tuning a pre-trained neural networks or the use of domain-specific hand-engineered features. Here we take the approach of treating network layers, or entire networks, as modules and combine pre-trained modules with untrained modules, to learn the shift in distributions between data sets. The central impact of using a modular approach comes from adding new representations to a network, as opposed to replacing representations via fine-tuning.more » Using this technique, we are able surpass results using standard fine-tuning transfer learning approaches, and we are also able to significantly increase performance over such approaches when using smaller amounts of data.« less

Color encoding in biologically-inspired convolutional neural networks.

PubMed

Rafegas, Ivet; Vanrell, Maria

2018-05-11

Convolutional Neural Networks have been proposed as suitable frameworks to model biological vision. Some of these artificial networks showed representational properties that rival primate performances in object recognition. In this paper we explore how color is encoded in a trained artificial network. It is performed by estimating a color selectivity index for each neuron, which allows us to describe the neuron activity to a color input stimuli. The index allows us to classify whether they are color selective or not and if they are of a single or double color. We have determined that all five convolutional layers of the network have a large number of color selective neurons. Color opponency clearly emerges in the first layer, presenting 4 main axes (Black-White, Red-Cyan, Blue-Yellow and Magenta-Green), but this is reduced and rotated as we go deeper into the network. In layer 2 we find a denser hue sampling of color neurons and opponency is reduced almost to one new main axis, the Bluish-Orangish coinciding with the dataset bias. In layers 3, 4 and 5 color neurons are similar amongst themselves, presenting different type of neurons that detect specific colored objects (e.g., orangish faces), specific surrounds (e.g., blue sky) or specific colored or contrasted object-surround configurations (e.g. blue blob in a green surround). Overall, our work concludes that color and shape representation are successively entangled through all the layers of the studied network, revealing certain parallelisms with the reported evidences in primate brains that can provide useful insight into intermediate hierarchical spatio-chromatic representations. Copyright © 2018 Elsevier Ltd. All rights reserved.

Comparisons of neural networks to standard techniques for image classification and correlation

NASA Technical Reports Server (NTRS)

Paola, Justin D.; Schowengerdt, Robert A.

1994-01-01

Neural network techniques for multispectral image classification and spatial pattern detection are compared to the standard techniques of maximum-likelihood classification and spatial correlation. The neural network produced a more accurate classification than maximum-likelihood of a Landsat scene of Tucson, Arizona. Some of the errors in the maximum-likelihood classification are illustrated using decision region and class probability density plots. As expected, the main drawback to the neural network method is the long time required for the training stage. The network was trained using several different hidden layer sizes to optimize both the classification accuracy and training speed, and it was found that one node per class was optimal. The performance improved when 3x3 local windows of image data were entered into the net. This modification introduces texture into the classification without explicit calculation of a texture measure. Larger windows were successfully used for the detection of spatial features in Landsat and Magellan synthetic aperture radar imagery.

Learning vector quantization neural networks improve accuracy of transcranial color-coded duplex sonography in detection of middle cerebral artery spasm--preliminary report.

PubMed

Swiercz, Miroslaw; Kochanowicz, Jan; Weigele, John; Hurst, Robert; Liebeskind, David S; Mariak, Zenon; Melhem, Elias R; Krejza, Jaroslaw

2008-01-01

To determine the performance of an artificial neural network in transcranial color-coded duplex sonography (TCCS) diagnosis of middle cerebral artery (MCA) spasm. TCCS was prospectively acquired within 2 h prior to routine cerebral angiography in 100 consecutive patients (54M:46F, median age 50 years). Angiographic MCA vasospasm was classified as mild (<25% of vessel caliber reduction), moderate (25-50%), or severe (>50%). A Learning Vector Quantization neural network classified MCA spasm based on TCCS peak-systolic, mean, and end-diastolic velocity data. During a four-class discrimination task, accurate classification by the network ranged from 64.9% to 72.3%, depending on the number of neurons in the Kohonen layer. Accurate classification of vasospasm ranged from 79.6% to 87.6%, with an accuracy of 84.7% to 92.1% for the detection of moderate-to-severe vasospasm. An artificial neural network may increase the accuracy of TCCS in diagnosis of MCA spasm.

Evolving RBF neural networks for adaptive soft-sensor design.

PubMed

Alexandridis, Alex

2013-12-01

This work presents an adaptive framework for building soft-sensors based on radial basis function (RBF) neural network models. The adaptive fuzzy means algorithm is utilized in order to evolve an RBF network, which approximates the unknown system based on input-output data from it. The methodology gradually builds the RBF network model, based on two separate levels of adaptation: On the first level, the structure of the hidden layer is modified by adding or deleting RBF centers, while on the second level, the synaptic weights are adjusted with the recursive least squares with exponential forgetting algorithm. The proposed approach is tested on two different systems, namely a simulated nonlinear DC Motor and a real industrial reactor. The results show that the produced soft-sensors can be successfully applied to model the two nonlinear systems. A comparison with two different adaptive modeling techniques, namely a dynamic evolving neural-fuzzy inference system (DENFIS) and neural networks trained with online backpropagation, highlights the advantages of the proposed methodology.

What the success of brain imaging implies about the neural code.

PubMed

Guest, Olivia; Love, Bradley C

2017-01-19

The success of fMRI places constraints on the nature of the neural code. The fact that researchers can infer similarities between neural representations, despite fMRI's limitations, implies that certain neural coding schemes are more likely than others. For fMRI to succeed given its low temporal and spatial resolution, the neural code must be smooth at the voxel and functional level such that similar stimuli engender similar internal representations. Through proof and simulation, we determine which coding schemes are plausible given both fMRI's successes and its limitations in measuring neural activity. Deep neural network approaches, which have been forwarded as computational accounts of the ventral stream, are consistent with the success of fMRI, though functional smoothness breaks down in the later network layers. These results have implications for the nature of the neural code and ventral stream, as well as what can be successfully investigated with fMRI.

Analog Delta-Back-Propagation Neural-Network Circuitry

NASA Technical Reports Server (NTRS)

Eberhart, Silvio

1990-01-01

Changes in synapse weights due to circuit drifts suppressed. Proposed fully parallel analog version of electronic neural-network processor based on delta-back-propagation algorithm. Processor able to "learn" when provided with suitable combinations of inputs and enforced outputs. Includes programmable resistive memory elements (corresponding to synapses), conductances (synapse weights) adjusted during learning. Buffer amplifiers, summing circuits, and sample-and-hold circuits arranged in layers of electronic neurons in accordance with delta-back-propagation algorithm.

Deep neural network convolution (NNC) for three-class classification of diffuse lung disease opacities in high-resolution CT (HRCT): consolidation, ground-glass opacity (GGO), and normal opacity

NASA Astrophysics Data System (ADS)

Hashimoto, Noriaki; Suzuki, Kenji; Liu, Junchi; Hirano, Yasushi; MacMahon, Heber; Kido, Shoji

2018-02-01

Consolidation and ground-glass opacity (GGO) are two major types of opacities associated with diffuse lung diseases. Accurate detection and classification of such opacities are crucially important in the diagnosis of lung diseases, but the process is subjective, and suffers from interobserver variability. Our study purpose was to develop a deep neural network convolution (NNC) system for distinguishing among consolidation, GGO, and normal lung tissue in high-resolution CT (HRCT). We developed ensemble of two deep NNC models, each of which was composed of neural network regression (NNR) with an input layer, a convolution layer, a fully-connected hidden layer, and a fully-connected output layer followed by a thresholding layer. The output layer of each NNC provided a map for the likelihood of being each corresponding lung opacity of interest. The two NNC models in the ensemble were connected in a class-selection layer. We trained our NNC ensemble with pairs of input 2D axial slices and "teaching" probability maps for the corresponding lung opacity, which were obtained by combining three radiologists' annotations. We randomly selected 10 and 40 slices from HRCT scans of 172 patients for each class as a training and test set, respectively. Our NNC ensemble achieved an area under the receiver-operating-characteristic (ROC) curve (AUC) of 0.981 and 0.958 in distinction of consolidation and GGO, respectively, from normal opacity, yielding a classification accuracy of 93.3% among 3 classes. Thus, our deep-NNC-based system for classifying diffuse lung diseases achieved high accuracies for classification of consolidation, GGO, and normal opacity.

Nonlinear inversion of electrical resistivity imaging using pruning Bayesian neural networks

NASA Astrophysics Data System (ADS)

Jiang, Fei-Bo; Dai, Qian-Wei; Dong, Li

2016-06-01

Conventional artificial neural networks used to solve electrical resistivity imaging (ERI) inversion problem suffer from overfitting and local minima. To solve these problems, we propose to use a pruning Bayesian neural network (PBNN) nonlinear inversion method and a sample design method based on the K-medoids clustering algorithm. In the sample design method, the training samples of the neural network are designed according to the prior information provided by the K-medoids clustering results; thus, the training process of the neural network is well guided. The proposed PBNN, based on Bayesian regularization, is used to select the hidden layer structure by assessing the effect of each hidden neuron to the inversion results. Then, the hyperparameter α k , which is based on the generalized mean, is chosen to guide the pruning process according to the prior distribution of the training samples under the small-sample condition. The proposed algorithm is more efficient than other common adaptive regularization methods in geophysics. The inversion of synthetic data and field data suggests that the proposed method suppresses the noise in the neural network training stage and enhances the generalization. The inversion results with the proposed method are better than those of the BPNN, RBFNN, and RRBFNN inversion methods as well as the conventional least squares inversion.

Expanding the occupational health methodology: A concatenated artificial neural network approach to model the burnout process in Chinese nurses.

PubMed

Ladstätter, Felix; Garrosa, Eva; Moreno-Jiménez, Bernardo; Ponsoda, Vicente; Reales Aviles, José Manuel; Dai, Junming

2016-01-01

Artificial neural networks are sophisticated modelling and prediction tools capable of extracting complex, non-linear relationships between predictor (input) and predicted (output) variables. This study explores this capacity by modelling non-linearities in the hardiness-modulated burnout process with a neural network. Specifically, two multi-layer feed-forward artificial neural networks are concatenated in an attempt to model the composite non-linear burnout process. Sensitivity analysis, a Monte Carlo-based global simulation technique, is then utilised to examine the first-order effects of the predictor variables on the burnout sub-dimensions and consequences. Results show that (1) this concatenated artificial neural network approach is feasible to model the burnout process, (2) sensitivity analysis is a prolific method to study the relative importance of predictor variables and (3) the relationships among variables involved in the development of burnout and its consequences are to different degrees non-linear. Many relationships among variables (e.g., stressors and strains) are not linear, yet researchers use linear methods such as Pearson correlation or linear regression to analyse these relationships. Artificial neural network analysis is an innovative method to analyse non-linear relationships and in combination with sensitivity analysis superior to linear methods.

Prediction of Flow Stress in Cadmium Using Constitutive Equation and Artificial Neural Network Approach

NASA Astrophysics Data System (ADS)

Sarkar, A.; Chakravartty, J. K.

2013-10-01

A model is developed to predict the constitutive flow behavior of cadmium during compression test using artificial neural network (ANN). The inputs of the neural network are strain, strain rate, and temperature, whereas flow stress is the output. Experimental data obtained from compression tests in the temperature range -30 to 70 °C, strain range 0.1 to 0.6, and strain rate range 10-3 to 1 s-1 are employed to develop the model. A three-layer feed-forward ANN is trained with Levenberg-Marquardt training algorithm. It has been shown that the developed ANN model can efficiently and accurately predict the deformation behavior of cadmium. This trained network could predict the flow stress better than a constitutive equation of the type.

Multi-layer neural networks for robot control

NASA Technical Reports Server (NTRS)

Pourboghrat, Farzad

1989-01-01

Two neural learning controller designs for manipulators are considered. The first design is based on a neural inverse-dynamics system. The second is the combination of the first one with a neural adaptive state feedback system. Both types of controllers enable the manipulator to perform any given task very well after a period of training and to do other untrained tasks satisfactorily. The second design also enables the manipulator to compensate for unpredictable perturbations.

Spike phase synchronization in multiplex cortical neural networks

NASA Astrophysics Data System (ADS)

Jalili, Mahdi

2017-01-01

In this paper we study synchronizability of two multiplex cortical networks: whole-cortex of hermaphrodite C. elegans and posterior cortex in male C. elegans. These networks are composed of two connection layers: network of chemical synapses and the one formed by gap junctions. This work studies the contribution of each layer on the phase synchronization of non-identical spiking Hindmarsh-Rose neurons. The network of male C. elegans shows higher phase synchronization than its randomized version, while it is not the case for hermaphrodite type. The random networks in each layer are constructed such that the nodes have the same degree as the original network, thus providing an unbiased comparison. In male C. elegans, although the gap junction network is sparser than the chemical network, it shows higher contribution in the synchronization phenomenon. This is not the case in hermaphrodite type, which is mainly due to significant less density of gap junction layer (0.013) as compared to chemical layer (0.028). Also, the gap junction network in this type has stronger community structure than the chemical network, and this is another driving factor for its weaker synchronizability.

Deep RNNs for video denoising

NASA Astrophysics Data System (ADS)

Chen, Xinyuan; Song, Li; Yang, Xiaokang

2016-09-01

Video denoising can be described as the problem of mapping from a specific length of noisy frames to clean one. We propose a deep architecture based on Recurrent Neural Network (RNN) for video denoising. The model learns a patch-based end-to-end mapping between the clean and noisy video sequences. It takes the corrupted video sequences as the input and outputs the clean one. Our deep network, which we refer to as deep Recurrent Neural Networks (deep RNNs or DRNNs), stacks RNN layers where each layer receives the hidden state of the previous layer as input. Experiment shows (i) the recurrent architecture through temporal domain extracts motion information and does favor to video denoising, and (ii) deep architecture have large enough capacity for expressing mapping relation between corrupted videos as input and clean videos as output, furthermore, (iii) the model has generality to learned different mappings from videos corrupted by different types of noise (e.g., Poisson-Gaussian noise). By training on large video databases, we are able to compete with some existing video denoising methods.

Design of Neural Networks for Fast Convergence and Accuracy: Dynamics and Control

NASA Technical Reports Server (NTRS)

Maghami, Peiman G.; Sparks, Dean W., Jr.

1997-01-01

A procedure for the design and training of artificial neural networks, used for rapid and efficient controls and dynamics design and analysis for flexible space systems, has been developed. Artificial neural networks are employed, such that once properly trained, they provide a means of evaluating the impact of design changes rapidly. Specifically, two-layer feedforward neural networks are designed to approximate the functional relationship between the component/spacecraft design changes and measures of its performance or nonlinear dynamics of the system/components. A training algorithm, based on statistical sampling theory, is presented, which guarantees that the trained networks provide a designer-specified degree of accuracy in mapping the functional relationship. Within each iteration of this statistical-based algorithm, a sequential design algorithm is used for the design and training of the feedforward network to provide rapid convergence to the network goals. Here, at each sequence a new network is trained to minimize the error of previous network. The proposed method should work for applications wherein an arbitrary large source of training data can be generated. Two numerical examples are performed on a spacecraft application in order to demonstrate the feasibility of the proposed approach.

Design of neural networks for fast convergence and accuracy: dynamics and control.

PubMed

Maghami, P G; Sparks, D R

2000-01-01

A procedure for the design and training of artificial neural networks, used for rapid and efficient controls and dynamics design and analysis for flexible space systems, has been developed. Artificial neural networks are employed, such that once properly trained, they provide a means of evaluating the impact of design changes rapidly. Specifically, two-layer feedforward neural networks are designed to approximate the functional relationship between the component/spacecraft design changes and measures of its performance or nonlinear dynamics of the system/components. A training algorithm, based on statistical sampling theory, is presented, which guarantees that the trained networks provide a designer-specified degree of accuracy in mapping the functional relationship. Within each iteration of this statistical-based algorithm, a sequential design algorithm is used for the design and training of the feedforward network to provide rapid convergence to the network goals. Here, at each sequence a new network is trained to minimize the error of previous network. The proposed method should work for applications wherein an arbitrary large source of training data can be generated. Two numerical examples are performed on a spacecraft application in order to demonstrate the feasibility of the proposed approach.

[Application of artificial neural networks on the prediction of surface ozone concentrations].

PubMed

Shen, Lu-Lu; Wang, Yu-Xuan; Duan, Lei

2011-08-01

Ozone is an important secondary air pollutant in the lower atmosphere. In order to predict the hourly maximum ozone one day in advance based on the meteorological variables for the Wanqingsha site in Guangzhou, Guangdong province, a neural network model (Multi-Layer Perceptron) and a multiple linear regression model were used and compared. Model inputs are meteorological parameters (wind speed, wind direction, air temperature, relative humidity, barometric pressure and solar radiation) of the next day and hourly maximum ozone concentration of the previous day. The OBS (optimal brain surgeon) was adopted to prune the neutral work, to reduce its complexity and to improve its generalization ability. We find that the pruned neural network has the capacity to predict the peak ozone, with an agreement index of 92.3%, the root mean square error of 0.0428 mg/m3, the R-square of 0.737 and the success index of threshold exceedance 77.0% (the threshold O3 mixing ratio of 0.20 mg/m3). When the neural classifier was added to the neural network model, the success index of threshold exceedance increased to 83.6%. Through comparison of the performance indices between the multiple linear regression model and the neural network model, we conclud that that neural network is a better choice to predict peak ozone from meteorological forecast, which may be applied to practical prediction of ozone concentration.

Three-D Artificial Neural Network (3DANN) technology. Blueprint for the future

NASA Technical Reports Server (NTRS)

Carson, John

1994-01-01

Irvine Sensors Corporation (ISC), working closely with JPL under BMDO/ONR sponsorship, is developing a radically new neural computing technology. Primarily aimed at discrimination and target recognition for BMDO missile interceptor applications, it appears to have near term commercial applicability to such problems as handwriting and face recognition, just to name two. In its earliest form it will be able to perform inner product computation using 262 thousand 64x64 templates (weighted synapse arrays) where the 64(exp 5) weights can all be changed every millisecond. Internal switching provides an inherent capability to zoom, translate, or rotate the templates. The 3D silicon architecture is manufactured on a commercial, high volume DRAM production line at very low cost, enabling its commercialization. Two technology thrusts are beginning: in the first, the 64 layer capability of 3DANN-I will be extended to 1024 layers and beyond. In the second layer size will be shrunk to 2-3 millimeters to reduce layer costs. Our workshop goal is to expose this technology to the neural network community as an emerging tool for their use and to obtain their desire for its future development.

Artificial neural networks in Space Station optimal attitude control

NASA Astrophysics Data System (ADS)

Kumar, Renjith R.; Seywald, Hans; Deshpande, Samir M.; Rahman, Zia

1995-01-01

Innovative techniques of using "artificial neural networks" (ANN) for improving the performance of the pitch axis attitude control system of Space Station Freedom using control moment gyros (CMGs) are investigated. The first technique uses a feed-forward ANN with multi-layer perceptrons to obtain an on-line controller which improves the performance of the control system via a model following approach. The second technique uses a single layer feed-forward ANN with a modified back propagation scheme to estimate the internal plant variations and the external disturbances separately. These estimates are then used to solve two differential Riccati equations to obtain time varying gains which improve the control system performance in successive orbits.

Understanding the neural basis of cognitive bias modification as a clinical treatment for depression.

PubMed

Eguchi, Akihiro; Walters, Daniel; Peerenboom, Nele; Dury, Hannah; Fox, Elaine; Stringer, Simon

2017-03-01

[Correction Notice: An Erratum for this article was reported in Vol 85(3) of Journal of Consulting and Clinical Psychology (see record 2017-07144-002). In the article, there was an error in the Discussion section's first paragraph for Implications and Future Work. The in-text reference citation for Penton-Voak et al. (2013) was incorrectly listed as "Blumenfeld, Preminger, Sagi, and Tsodyks (2006)". All versions of this article have been corrected.] Objective: Cognitive bias modification (CBM) eliminates cognitive biases toward negative information and is efficacious in reducing depression recurrence, but the mechanisms behind the bias elimination are not fully understood. The present study investigated, through computer simulation of neural network models, the neural dynamics underlying the use of CBM in eliminating the negative biases in the way that depressed patients evaluate facial expressions. We investigated 2 new CBM methodologies using biologically plausible synaptic learning mechanisms-continuous transformation learning and trace learning-which guide learning by exploiting either the spatial or temporal continuity between visual stimuli presented during training. We first describe simulations with a simplified 1-layer neural network, and then we describe simulations in a biologically detailed multilayer neural network model of the ventral visual pathway. After training with either the continuous transformation learning rule or the trace learning rule, the 1-layer neural network eliminated biases in interpreting neutral stimuli as sad. The multilayer neural network trained with realistic face stimuli was also shown to be able to use continuous transformation learning or trace learning to reduce biases in the interpretation of neutral stimuli. The simulation results suggest 2 biologically plausible synaptic learning mechanisms, continuous transformation learning and trace learning, that may subserve CBM. The results are highly informative for the development of experimental protocols to produce optimal CBM training methodologies with human participants. (PsycINFO Database Record (c) 2017 APA, all rights reserved).

Artificial vision by multi-layered neural networks: neocognitron and its advances.

PubMed

Fukushima, Kunihiko

2013-01-01

The neocognitron is a neural network model proposed by Fukushima (1980). Its architecture was suggested by neurophysiological findings on the visual systems of mammals. It is a hierarchical multi-layered network. It acquires the ability to robustly recognize visual patterns through learning. Although the neocognitron has a long history, modifications of the network to improve its performance are still going on. For example, a recent neocognitron uses a new learning rule, named add-if-silent, which makes the learning process much simpler and more stable. Nevertheless, a high recognition rate can be kept with a smaller scale of the network. Referring to the history of the neocognitron, this paper discusses recent advances in the neocognitron. We also show that various new functions can be realized by, for example, introducing top-down connections to the neocognitron: mechanism of selective attention, recognition and completion of partly occluded patterns, restoring occluded contours, and so on. Copyright © 2012 Elsevier Ltd. All rights reserved.

Computational models of location-invariant orthographic processing

NASA Astrophysics Data System (ADS)

Dandurand, Frédéric; Hannagan, Thomas; Grainger, Jonathan

2013-03-01

We trained three topologies of backpropagation neural networks to discriminate 2000 words (lexical representations) presented at different positions of a horizontal letter array. The first topology (zero-deck) contains no hidden layer, the second (one-deck) has a single hidden layer, and for the last topology (two-deck), the task is divided in two subtasks implemented as two stacked neural networks, with explicit word-centred letters as intermediate representations. All topologies successfully simulated two key benchmark phenomena observed in skilled human reading: transposed-letter priming and relative-position priming. However, the two-deck topology most accurately simulated the ability to discriminate words from nonwords, while containing the fewest connection weights. We analysed the internal representations after training. Zero-deck networks implement a letter-based scheme with a position bias to differentiate anagrams. One-deck networks implement a holographic overlap coding in which representations are essentially letter-based and words are linear combinations of letters. Two-deck networks also implement holographic-coding.

Gamma Spectroscopy by Artificial Neural Network Coupled with MCNP

NASA Astrophysics Data System (ADS)

Sahiner, Huseyin

While neutron activation analysis is widely used in many areas, sensitivity of the analysis depends on how the analysis is conducted. Even though the sensitivity of the techniques carries error, compared to chemical analysis, its range is in parts per million or sometimes billion. Due to this sensitivity, the use of neutron activation analysis becomes important when analyzing bio-samples. Artificial neural network is an attractive technique for complex systems. Although there are neural network applications on spectral analysis, training by simulated data to analyze experimental data has not been made. This study offers an improvement on spectral analysis and optimization on neural network for the purpose. The work considers five elements that are considered as trace elements for bio-samples. However, the system is not limited to five elements. The only limitation of the study comes from data library availability on MCNP. A perceptron network was employed to identify five elements from gamma spectra. In quantitative analysis, better results were obtained when the neural fitting tool in MATLAB was used. As a training function, Levenberg-Marquardt algorithm was used with 23 neurons in the hidden layer with 259 gamma spectra in the input. Because the interest of the study deals with five elements, five neurons representing peak counts of five isotopes in the input layer were used. Five output neurons revealed mass information of these elements from irradiated kidney stones. Results showing max error of 17.9% in APA, 24.9% in UA, 28.2% in COM, 27.9% in STRU type showed the success of neural network approach in analyzing gamma spectra. This high error was attributed to Zn that has a very long decay half-life compared to the other elements. The simulation and experiments were made under certain experimental setup (3 hours irradiation, 96 hours decay time, 8 hours counting time). Nevertheless, the approach is subject to be generalized for different setups.

A biologically inspired neural net for trajectory formation and obstacle avoidance.

PubMed

Glasius, R; Komoda, A; Gielen, S C

1996-06-01

In this paper we present a biologically inspired two-layered neural network for trajectory formation and obstacle avoidance. The two topographically ordered neural maps consist of analog neurons having continuous dynamics. The first layer, the sensory map, receives sensory information and builds up an activity pattern which contains the optimal solution (i.e. shortest path without collisions) for any given set of current position, target positions and obstacle positions. Targets and obstacles are allowed to move, in which case the activity pattern in the sensory map will change accordingly. The time evolution of the neural activity in the second layer, the motor map, results in a moving cluster of activity, which can be interpreted as a population vector. Through the feedforward connections between the two layers, input of the sensory map directs the movement of the cluster along the optimal path from the current position of the cluster to the target position. The smooth trajectory is the result of the intrinsic dynamics of the network only. No supervisor is required. The output of the motor map can be used for direct control of an autonomous system in a cluttered environment or for control of the actuators of a biological limb or robot manipulator. The system is able to reach a target even in the presence of an external perturbation. Computer simulations of a point robot and a multi-joint manipulator illustrate the theory.

Neuronal avalanches of a self-organized neural network with active-neuron-dominant structure.

PubMed

Li, Xiumin; Small, Michael

2012-06-01

Neuronal avalanche is a spontaneous neuronal activity which obeys a power-law distribution of population event sizes with an exponent of -3/2. It has been observed in the superficial layers of cortex both in vivo and in vitro. In this paper, we analyze the information transmission of a novel self-organized neural network with active-neuron-dominant structure. Neuronal avalanches can be observed in this network with appropriate input intensity. We find that the process of network learning via spike-timing dependent plasticity dramatically increases the complexity of network structure, which is finally self-organized to be active-neuron-dominant connectivity. Both the entropy of activity patterns and the complexity of their resulting post-synaptic inputs are maximized when the network dynamics are propagated as neuronal avalanches. This emergent topology is beneficial for information transmission with high efficiency and also could be responsible for the large information capacity of this network compared with alternative archetypal networks with different neural connectivity.

APPLICATION OF NEURAL NETWORK ALGORITHMS FOR BPM LINEARIZATION

DOE Office of Scientific and Technical Information (OSTI.GOV)

Musson, John C.; Seaton, Chad; Spata, Mike F.

2012-11-01

Stripline BPM sensors contain inherent non-linearities, as a result of field distortions from the pickup elements. Many methods have been devised to facilitate corrections, often employing polynomial fitting. The cost of computation makes real-time correction difficult, particulalry when integer math is utilized. The application of neural-network technology, particularly the multi-layer perceptron algorithm, is proposed as an efficient alternative for electrode linearization. A process of supervised learning is initially used to determine the weighting coefficients, which are subsequently applied to the incoming electrode data. A non-linear layer, known as an activation layer, is responsible for the removal of saturation effects. Implementationmore » of a perceptron in an FPGA-based software-defined radio (SDR) is presented, along with performance comparisons. In addition, efficient calculation of the sigmoidal activation function via the CORDIC algorithm is presented.« less

An Event-Driven Classifier for Spiking Neural Networks Fed with Synthetic or Dynamic Vision Sensor Data.

PubMed

Stromatias, Evangelos; Soto, Miguel; Serrano-Gotarredona, Teresa; Linares-Barranco, Bernabé

2017-01-01

This paper introduces a novel methodology for training an event-driven classifier within a Spiking Neural Network (SNN) System capable of yielding good classification results when using both synthetic input data and real data captured from Dynamic Vision Sensor (DVS) chips. The proposed supervised method uses the spiking activity provided by an arbitrary topology of prior SNN layers to build histograms and train the classifier in the frame domain using the stochastic gradient descent algorithm. In addition, this approach can cope with leaky integrate-and-fire neuron models within the SNN, a desirable feature for real-world SNN applications, where neural activation must fade away after some time in the absence of inputs. Consequently, this way of building histograms captures the dynamics of spikes immediately before the classifier. We tested our method on the MNIST data set using different synthetic encodings and real DVS sensory data sets such as N-MNIST, MNIST-DVS, and Poker-DVS using the same network topology and feature maps. We demonstrate the effectiveness of our approach by achieving the highest classification accuracy reported on the N-MNIST (97.77%) and Poker-DVS (100%) real DVS data sets to date with a spiking convolutional network. Moreover, by using the proposed method we were able to retrain the output layer of a previously reported spiking neural network and increase its performance by 2%, suggesting that the proposed classifier can be used as the output layer in works where features are extracted using unsupervised spike-based learning methods. In addition, we also analyze SNN performance figures such as total event activity and network latencies, which are relevant for eventual hardware implementations. In summary, the paper aggregates unsupervised-trained SNNs with a supervised-trained SNN classifier, combining and applying them to heterogeneous sets of benchmarks, both synthetic and from real DVS chips.

Optimization of multilayer neural network parameters for speaker recognition

NASA Astrophysics Data System (ADS)

Tovarek, Jaromir; Partila, Pavol; Rozhon, Jan; Voznak, Miroslav; Skapa, Jan; Uhrin, Dominik; Chmelikova, Zdenka

2016-05-01

This article discusses the impact of multilayer neural network parameters for speaker identification. The main task of speaker identification is to find a specific person in the known set of speakers. It means that the voice of an unknown speaker (wanted person) belongs to a group of reference speakers from the voice database. One of the requests was to develop the text-independent system, which means to classify wanted person regardless of content and language. Multilayer neural network has been used for speaker identification in this research. Artificial neural network (ANN) needs to set parameters like activation function of neurons, steepness of activation functions, learning rate, the maximum number of iterations and a number of neurons in the hidden and output layers. ANN accuracy and validation time are directly influenced by the parameter settings. Different roles require different settings. Identification accuracy and ANN validation time were evaluated with the same input data but different parameter settings. The goal was to find parameters for the neural network with the highest precision and shortest validation time. Input data of neural networks are a Mel-frequency cepstral coefficients (MFCC). These parameters describe the properties of the vocal tract. Audio samples were recorded for all speakers in a laboratory environment. Training, testing and validation data set were split into 70, 15 and 15 %. The result of the research described in this article is different parameter setting for the multilayer neural network for four speakers.

Complex Rotation Quantum Dynamic Neural Networks (CRQDNN) using Complex Quantum Neuron (CQN): Applications to time series prediction.

PubMed

Cui, Yiqian; Shi, Junyou; Wang, Zili

2015-11-01

Quantum Neural Networks (QNN) models have attracted great attention since it innovates a new neural computing manner based on quantum entanglement. However, the existing QNN models are mainly based on the real quantum operations, and the potential of quantum entanglement is not fully exploited. In this paper, we proposes a novel quantum neuron model called Complex Quantum Neuron (CQN) that realizes a deep quantum entanglement. Also, a novel hybrid networks model Complex Rotation Quantum Dynamic Neural Networks (CRQDNN) is proposed based on Complex Quantum Neuron (CQN). CRQDNN is a three layer model with both CQN and classical neurons. An infinite impulse response (IIR) filter is embedded in the Networks model to enable the memory function to process time series inputs. The Levenberg-Marquardt (LM) algorithm is used for fast parameter learning. The networks model is developed to conduct time series predictions. Two application studies are done in this paper, including the chaotic time series prediction and electronic remaining useful life (RUL) prediction. Copyright © 2015 Elsevier Ltd. All rights reserved.

A Spatiotemporal Prediction Framework for Air Pollution Based on Deep RNN

NASA Astrophysics Data System (ADS)

Fan, J.; Li, Q.; Hou, J.; Feng, X.; Karimian, H.; Lin, S.

2017-10-01

Time series data in practical applications always contain missing values due to sensor malfunction, network failure, outliers etc. In order to handle missing values in time series, as well as the lack of considering temporal properties in machine learning models, we propose a spatiotemporal prediction framework based on missing value processing algorithms and deep recurrent neural network (DRNN). By using missing tag and missing interval to represent time series patterns, we implement three different missing value fixing algorithms, which are further incorporated into deep neural network that consists of LSTM (Long Short-term Memory) layers and fully connected layers. Real-world air quality and meteorological datasets (Jingjinji area, China) are used for model training and testing. Deep feed forward neural networks (DFNN) and gradient boosting decision trees (GBDT) are trained as baseline models against the proposed DRNN. Performances of three missing value fixing algorithms, as well as different machine learning models are evaluated and analysed. Experiments show that the proposed DRNN framework outperforms both DFNN and GBDT, therefore validating the capacity of the proposed framework. Our results also provides useful insights for better understanding of different strategies that handle missing values.

On-line training of recurrent neural networks with continuous topology adaptation.

PubMed

Obradovic, D

1996-01-01

This paper presents an online procedure for training dynamic neural networks with input-output recurrences whose topology is continuously adjusted to the complexity of the target system dynamics. This is accomplished by changing the number of the elements of the network hidden layer whenever the existing topology cannot capture the dynamics presented by the new data. The training mechanism is based on the suitably altered extended Kalman filter (EKF) algorithm which is simultaneously used for the network parameter adjustment and for its state estimation. The network consists of a single hidden layer with Gaussian radial basis functions (GRBF), and a linear output layer. The choice of the GRBF is induced by the requirements of the online learning. The latter implies the network architecture which permits only local influence of the new data point in order not to forget the previously learned dynamics. The continuous topology adaptation is implemented in our algorithm to avoid memory and computational problems of using a regular grid of GRBF'S which covers the network input space. Furthermore, we show that the resulting parameter increase can be handled "smoothly" without interfering with the already acquired information. If the target system dynamics are changing over time, we show that a suitable forgetting factor can be used to "unlearn" the no longer-relevant dynamics. The quality of the recurrent network training algorithm is demonstrated on the identification of nonlinear dynamic systems.

Information processing in echo state networks at the edge of chaos.

PubMed

Boedecker, Joschka; Obst, Oliver; Lizier, Joseph T; Mayer, N Michael; Asada, Minoru

2012-09-01

We investigate information processing in randomly connected recurrent neural networks. It has been shown previously that the computational capabilities of these networks are maximized when the recurrent layer is close to the border between a stable and an unstable dynamics regime, the so called edge of chaos. The reasons, however, for this maximized performance are not completely understood. We adopt an information-theoretical framework and are for the first time able to quantify the computational capabilities between elements of these networks directly as they undergo the phase transition to chaos. Specifically, we present evidence that both information transfer and storage in the recurrent layer are maximized close to this phase transition, providing an explanation for why guiding the recurrent layer toward the edge of chaos is computationally useful. As a consequence, our study suggests self-organized ways of improving performance in recurrent neural networks, driven by input data. Moreover, the networks we study share important features with biological systems such as feedback connections and online computation on input streams. A key example is the cerebral cortex, which was shown to also operate close to the edge of chaos. Consequently, the behavior of model systems as studied here is likely to shed light on reasons why biological systems are tuned into this specific regime.

Spatial frequency domain spectroscopy of two layer media

NASA Astrophysics Data System (ADS)

Yudovsky, Dmitry; Durkin, Anthony J.

2011-10-01

Monitoring of tissue blood volume and oxygen saturation using biomedical optics techniques has the potential to inform the assessment of tissue health, healing, and dysfunction. These quantities are typically estimated from the contribution of oxyhemoglobin and deoxyhemoglobin to the absorption spectrum of the dermis. However, estimation of blood related absorption in superficial tissue such as the skin can be confounded by the strong absorption of melanin in the epidermis. Furthermore, epidermal thickness and pigmentation varies with anatomic location, race, gender, and degree of disease progression. This study describes a technique for decoupling the effect of melanin absorption in the epidermis from blood absorption in the dermis for a large range of skin types and thicknesses. An artificial neural network was used to map input optical properties to spatial frequency domain diffuse reflectance of two layer media. Then, iterative fitting was used to determine the optical properties from simulated spatial frequency domain diffuse reflectance. Additionally, an artificial neural network was trained to directly map spatial frequency domain reflectance to sets of optical properties of a two layer medium, thus bypassing the need for iteration. In both cases, the optical thickness of the epidermis and absorption and reduced scattering coefficients of the dermis were determined independently. The accuracy and efficiency of the iterative fitting approach was compared with the direct neural network inversion.

Supervised artificial neural network-based method for conversion of solar radiation data (case study: Algeria)

NASA Astrophysics Data System (ADS)

Laidi, Maamar; Hanini, Salah; Rezrazi, Ahmed; Yaiche, Mohamed Redha; El Hadj, Abdallah Abdallah; Chellali, Farouk

2017-04-01

In this study, a backpropagation artificial neural network (BP-ANN) model is used as an alternative approach to predict solar radiation on tilted surfaces (SRT) using a number of variables involved in physical process. These variables are namely the latitude of the site, mean temperature and relative humidity, Linke turbidity factor and Angstrom coefficient, extraterrestrial solar radiation, solar radiation data measured on horizontal surfaces (SRH), and solar zenith angle. Experimental solar radiation data from 13 stations spread all over Algeria around the year (2004) were used for training/validation and testing the artificial neural networks (ANNs), and one station was used to make the interpolation of the designed ANN. The ANN model was trained, validated, and tested using 60, 20, and 20 % of all data, respectively. The configuration 8-35-1 (8 inputs, 35 hidden, and 1 output neurons) presented an excellent agreement between the prediction and the experimental data during the test stage with determination coefficient of 0.99 and root meat squared error of 5.75 Wh/m2, considering a three-layer feedforward backpropagation neural network with Levenberg-Marquardt training algorithm, a hyperbolic tangent sigmoid and linear transfer function at the hidden and the output layer, respectively. This novel model could be used by researchers or scientists to design high-efficiency solar devices that are usually tilted at an optimum angle to increase the solar incident on the surface.

Residual Deep Convolutional Neural Network Predicts MGMT Methylation Status.

PubMed

Korfiatis, Panagiotis; Kline, Timothy L; Lachance, Daniel H; Parney, Ian F; Buckner, Jan C; Erickson, Bradley J

2017-10-01

Predicting methylation of the O6-methylguanine methyltransferase (MGMT) gene status utilizing MRI imaging is of high importance since it is a predictor of response and prognosis in brain tumors. In this study, we compare three different residual deep neural network (ResNet) architectures to evaluate their ability in predicting MGMT methylation status without the need for a distinct tumor segmentation step. We found that the ResNet50 (50 layers) architecture was the best performing model, achieving an accuracy of 94.90% (+/- 3.92%) for the test set (classification of a slice as no tumor, methylated MGMT, or non-methylated). ResNet34 (34 layers) achieved 80.72% (+/- 13.61%) while ResNet18 (18 layers) accuracy was 76.75% (+/- 20.67%). ResNet50 performance was statistically significantly better than both ResNet18 and ResNet34 architectures (p < 0.001). We report a method that alleviates the need of extensive preprocessing and acts as a proof of concept that deep neural architectures can be used to predict molecular biomarkers from routine medical images.

Spatial interpolation and radiological mapping of ambient gamma dose rate by using artificial neural networks and fuzzy logic methods.

PubMed

Yeşilkanat, Cafer Mert; Kobya, Yaşar; Taşkın, Halim; Çevik, Uğur

2017-09-01

The aim of this study was to determine spatial risk dispersion of ambient gamma dose rate (AGDR) by using both artificial neural network (ANN) and fuzzy logic (FL) methods, compare the performances of methods, make dose estimations for intermediate stations with no previous measurements and create dose rate risk maps of the study area. In order to determine the dose distribution by using artificial neural networks, two main networks and five different network structures were used; feed forward ANN; Multi-layer perceptron (MLP), Radial basis functional neural network (RBFNN), Quantile regression neural network (QRNN) and recurrent ANN; Jordan networks (JN), Elman networks (EN). In the evaluation of estimation performance obtained for the test data, all models appear to give similar results. According to the cross-validation results obtained for explaining AGDR distribution, Pearson's r coefficients were calculated as 0.94, 0.91, 0.89, 0.91, 0.91 and 0.92 and RMSE values were calculated as 34.78, 43.28, 63.92, 44.86, 46.77 and 37.92 for MLP, RBFNN, QRNN, JN, EN and FL, respectively. In addition, spatial risk maps showing distributions of AGDR of the study area were created by all models and results were compared with geological, topological and soil structure. Copyright © 2017 Elsevier Ltd. All rights reserved.

Developing a Neural Network to Act as a Noise Filter

DTIC Science & Technology

1992-10-02

s... ... j.. .3885*I888+.. I. . . fff~8*I*8s+ ... i £888&aiz$$88 1. *.󈧧 z*I* sseI . .** m $as+.’ . . .𔄂seEgzsssas* ... .. +8891MI1$883< I. .. (*8...6: Architectures that provided the best results Using a Training Set of 5 Samples (no bias) Size of] m 1 ax 1 lumber Learning1 PatternI Overlap...Riddle IAbs I RfAS I of I Time Layer Layer Layer Error Error _ycles __(Seel_ m 5 x 4 (0,21 5 x 9 0.100 0.039 132 1043.0 lone of the neural netvork

Novel maximum-margin training algorithms for supervised neural networks.

PubMed

Ludwig, Oswaldo; Nunes, Urbano

2010-06-01

This paper proposes three novel training methods, two of them based on the backpropagation approach and a third one based on information theory for multilayer perceptron (MLP) binary classifiers. Both backpropagation methods are based on the maximal-margin (MM) principle. The first one, based on the gradient descent with adaptive learning rate algorithm (GDX) and named maximum-margin GDX (MMGDX), directly increases the margin of the MLP output-layer hyperplane. The proposed method jointly optimizes both MLP layers in a single process, backpropagating the gradient of an MM-based objective function, through the output and hidden layers, in order to create a hidden-layer space that enables a higher margin for the output-layer hyperplane, avoiding the testing of many arbitrary kernels, as occurs in case of support vector machine (SVM) training. The proposed MM-based objective function aims to stretch out the margin to its limit. An objective function based on Lp-norm is also proposed in order to take into account the idea of support vectors, however, overcoming the complexity involved in solving a constrained optimization problem, usually in SVM training. In fact, all the training methods proposed in this paper have time and space complexities O(N) while usual SVM training methods have time complexity O(N (3)) and space complexity O(N (2)) , where N is the training-data-set size. The second approach, named minimization of interclass interference (MICI), has an objective function inspired on the Fisher discriminant analysis. Such algorithm aims to create an MLP hidden output where the patterns have a desirable statistical distribution. In both training methods, the maximum area under ROC curve (AUC) is applied as stop criterion. The third approach offers a robust training framework able to take the best of each proposed training method. The main idea is to compose a neural model by using neurons extracted from three other neural networks, each one previously trained by MICI, MMGDX, and Levenberg-Marquard (LM), respectively. The resulting neural network was named assembled neural network (ASNN). Benchmark data sets of real-world problems have been used in experiments that enable a comparison with other state-of-the-art classifiers. The results provide evidence of the effectiveness of our methods regarding accuracy, AUC, and balanced error rate.

[Relationship between reference values of fibrinogen and geographical factors based on neural network analysis].

PubMed

Li, Meng-Jiao; Ge, Miao; Wang, Cong-Xia; Cen, Min-Yi; Jiang, Ji-Lin; He, Jin-Wei; Lin, Qian-Yi; Liu, Xin

2016-08-20

To analyze the relationship between the reference values of fibrinogen (FIB) in healthy Chinese adults and geographical factors to provide scientific evidences for establishing the uniform standard. The reference values of FIB of 10701 Chinese healthy adults from 103 cities were collected to investigate their relationship with 18 geographical factors including spatial index, terrain index, climate index, and soil index. Geographical factors that significantly correlated with the reference values were selected for constructing the BP neural network model. The spatial distribution map of the reference value of FIB of healthy Chinese adults was fitted by disjunctive kriging interpolation. We used the 5-layer neural network and selected 2000 times of training covering 11 hidden layers to build the simulation rule for simulating the relationship between FIB and geographical environmental factors using the MATLAB software. s The reference value of FIB in healthy Chinese adults was significantly correlated with the latitude, sunshine duration, annual average temperature, annual average relative humidity, annual precipitation, annual range of air temperature, average annual soil gravel content, and soil cation exchange capacity (silt). The artificial neural networks were created to analyze the simulation of the selected indicators of geographical factors. The spatial distribution map of the reference values of FIB in healthy Chinese adults showed a distribution pattern that FIB levels were higher in the South and lower in the North, and higher in the East and lower in the West. When the geographical factors of a certain area are known, the reference values of FIB in healthy Chinese adults can be obtained by establishing the neural network mode or plotting the spatial distribution map.

Shakeout: A New Approach to Regularized Deep Neural Network Training.

PubMed

Kang, Guoliang; Li, Jun; Tao, Dacheng

2018-05-01

Recent years have witnessed the success of deep neural networks in dealing with a plenty of practical problems. Dropout has played an essential role in many successful deep neural networks, by inducing regularization in the model training. In this paper, we present a new regularized training approach: Shakeout. Instead of randomly discarding units as Dropout does at the training stage, Shakeout randomly chooses to enhance or reverse each unit's contribution to the next layer. This minor modification of Dropout has the statistical trait: the regularizer induced by Shakeout adaptively combines , and regularization terms. Our classification experiments with representative deep architectures on image datasets MNIST, CIFAR-10 and ImageNet show that Shakeout deals with over-fitting effectively and outperforms Dropout. We empirically demonstrate that Shakeout leads to sparser weights under both unsupervised and supervised settings. Shakeout also leads to the grouping effect of the input units in a layer. Considering the weights in reflecting the importance of connections, Shakeout is superior to Dropout, which is valuable for the deep model compression. Moreover, we demonstrate that Shakeout can effectively reduce the instability of the training process of the deep architecture.

Statistical Mechanical Analysis of Online Learning with Weight Normalization in Single Layer Perceptron

NASA Astrophysics Data System (ADS)

Yoshida, Yuki; Karakida, Ryo; Okada, Masato; Amari, Shun-ichi

2017-04-01

Weight normalization, a newly proposed optimization method for neural networks by Salimans and Kingma (2016), decomposes the weight vector of a neural network into a radial length and a direction vector, and the decomposed parameters follow their steepest descent update. They reported that learning with the weight normalization achieves better converging speed in several tasks including image recognition and reinforcement learning than learning with the conventional parameterization. However, it remains theoretically uncovered how the weight normalization improves the converging speed. In this study, we applied a statistical mechanical technique to analyze on-line learning in single layer linear and nonlinear perceptrons with weight normalization. By deriving order parameters of the learning dynamics, we confirmed quantitatively that weight normalization realizes fast converging speed by automatically tuning the effective learning rate, regardless of the nonlinearity of the neural network. This property is realized when the initial value of the radial length is near the global minimum; therefore, our theory suggests that it is important to choose the initial value of the radial length appropriately when using weight normalization.

Mullite ceramic membranes for industrial oily wastewater treatment: experimental and neural network modeling.

PubMed

Shokrkar, H; Salahi, A; Kasiri, N; Mohammadi, T

2011-01-01

In this paper, results of an experimental and modeling of separation of oil from industrial oily wastewaters (desalter unit effluent of Seraje, Ghom gas wells, Iran) with mullite ceramic membranes are presented. Mullite microfiltration symmetric membranes were synthesized from kaolin clay and alpha-alumina powder. The results show that the mullite ceramic membrane has a high total organic carbon and chemical oxygen demand rejection (94 and 89%, respectively), a low fouling resistance (30%) and a high final permeation flux (75 L/m2 h). Also, an artificial neural network, a predictive tool for tracking the inputs and outputs of a non-linear problem, is used to model the permeation flux decline during microfiltration of oily wastewater. The aim was to predict the permeation flux as a function of feed temperature, trans-membrane pressure, cross-flow velocity, oil concentration and filtration time, using a feed-forward neural network. Finally the structure of hidden layers and nodes in each layer with minimum error were reported leading to a 4-15 structure which demonstrated good agreement with the experimental measurements with an average error of less than 2%.

An Efficient Supervised Training Algorithm for Multilayer Spiking Neural Networks

PubMed Central

Xie, Xiurui; Qu, Hong; Liu, Guisong; Zhang, Malu; Kurths, Jürgen

2016-01-01

The spiking neural networks (SNNs) are the third generation of neural networks and perform remarkably well in cognitive tasks such as pattern recognition. The spike emitting and information processing mechanisms found in biological cognitive systems motivate the application of the hierarchical structure and temporal encoding mechanism in spiking neural networks, which have exhibited strong computational capability. However, the hierarchical structure and temporal encoding approach require neurons to process information serially in space and time respectively, which reduce the training efficiency significantly. For training the hierarchical SNNs, most existing methods are based on the traditional back-propagation algorithm, inheriting its drawbacks of the gradient diffusion and the sensitivity on parameters. To keep the powerful computation capability of the hierarchical structure and temporal encoding mechanism, but to overcome the low efficiency of the existing algorithms, a new training algorithm, the Normalized Spiking Error Back Propagation (NSEBP) is proposed in this paper. In the feedforward calculation, the output spike times are calculated by solving the quadratic function in the spike response model instead of detecting postsynaptic voltage states at all time points in traditional algorithms. Besides, in the feedback weight modification, the computational error is propagated to previous layers by the presynaptic spike jitter instead of the gradient decent rule, which realizes the layer-wised training. Furthermore, our algorithm investigates the mathematical relation between the weight variation and voltage error change, which makes the normalization in the weight modification applicable. Adopting these strategies, our algorithm outperforms the traditional SNN multi-layer algorithms in terms of learning efficiency and parameter sensitivity, that are also demonstrated by the comprehensive experimental results in this paper. PMID:27044001

A model of traffic signs recognition with convolutional neural network

NASA Astrophysics Data System (ADS)

Hu, Haihe; Li, Yujian; Zhang, Ting; Huo, Yi; Kuang, Wenqing

2016-10-01

In real traffic scenes, the quality of captured images are generally low due to some factors such as lighting conditions, and occlusion on. All of these factors are challengeable for automated recognition algorithms of traffic signs. Deep learning has provided a new way to solve this kind of problems recently. The deep network can automatically learn features from a large number of data samples and obtain an excellent recognition performance. We therefore approach this task of recognition of traffic signs as a general vision problem, with few assumptions related to road signs. We propose a model of Convolutional Neural Network (CNN) and apply the model to the task of traffic signs recognition. The proposed model adopts deep CNN as the supervised learning model, directly takes the collected traffic signs image as the input, alternates the convolutional layer and subsampling layer, and automatically extracts the features for the recognition of the traffic signs images. The proposed model includes an input layer, three convolutional layers, three subsampling layers, a fully-connected layer, and an output layer. To validate the proposed model, the experiments are implemented using the public dataset of China competition of fuzzy image processing. Experimental results show that the proposed model produces a recognition accuracy of 99.01 % on the training dataset, and yield a record of 92% on the preliminary contest within the fourth best.

Speaker identification for the improvement of the security communication between law enforcement units

NASA Astrophysics Data System (ADS)

Tovarek, Jaromir; Partila, Pavol

2017-05-01

This article discusses the speaker identification for the improvement of the security communication between law enforcement units. The main task of this research was to develop the text-independent speaker identification system which can be used for real-time recognition. This system is designed for identification in the open set. It means that the unknown speaker can be anyone. Communication itself is secured, but we have to check the authorization of the communication parties. We have to decide if the unknown speaker is the authorized for the given action. The calls are recorded by IP telephony server and then these recordings are evaluate using classification If the system evaluates that the speaker is not authorized, it sends a warning message to the administrator. This message can detect, for example a stolen phone or other unusual situation. The administrator then performs the appropriate actions. Our novel proposal system uses multilayer neural network for classification and it consists of three layers (input layer, hidden layer, and output layer). A number of neurons in input layer corresponds with the length of speech features. Output layer then represents classified speakers. Artificial Neural Network classifies speech signal frame by frame, but the final decision is done over the complete record. This rule substantially increases accuracy of the classification. Input data for the neural network are a thirteen Mel-frequency cepstral coefficients, which describe the behavior of the vocal tract. These parameters are the most used for speaker recognition. Parameters for training, testing and validation were extracted from recordings of authorized users. Recording conditions for training data correspond with the real traffic of the system (sampling frequency, bit rate). The main benefit of the research is the system developed for text-independent speaker identification which is applied to secure communication between law enforcement units.

Prediction of strain values in reinforcements and concrete of a RC frame using neural networks

NASA Astrophysics Data System (ADS)

Vafaei, Mohammadreza; Alih, Sophia C.; Shad, Hossein; Falah, Ali; Halim, Nur Hajarul Falahi Abdul

2018-03-01

The level of strain in structural elements is an important indicator for the presence of damage and its intensity. Considering this fact, often structural health monitoring systems employ strain gauges to measure strains in critical elements. However, because of their sensitivity to the magnetic fields, inadequate long-term durability especially in harsh environments, difficulties in installation on existing structures, and maintenance cost, installation of strain gauges is not always possible for all structural components. Therefore, a reliable method that can accurately estimate strain values in critical structural elements is necessary for damage identification. In this study, a full-scale test was conducted on a planar RC frame to investigate the capability of neural networks for predicting the strain values. Two neural networks each of which having a single hidden layer was trained to relate the measured rotations and vertical displacements of the frame to the strain values measured at different locations of the frame. Results of trained neural networks indicated that they accurately estimated the strain values both in reinforcements and concrete. In addition, the trained neural networks were capable of predicting strains for the unseen input data set.

On the complexity of neural network classifiers: a comparison between shallow and deep architectures.

PubMed

Bianchini, Monica; Scarselli, Franco

2014-08-01

Recently, researchers in the artificial neural network field have focused their attention on connectionist models composed by several hidden layers. In fact, experimental results and heuristic considerations suggest that deep architectures are more suitable than shallow ones for modern applications, facing very complex problems, e.g., vision and human language understanding. However, the actual theoretical results supporting such a claim are still few and incomplete. In this paper, we propose a new approach to study how the depth of feedforward neural networks impacts on their ability in implementing high complexity functions. First, a new measure based on topological concepts is introduced, aimed at evaluating the complexity of the function implemented by a neural network, used for classification purposes. Then, deep and shallow neural architectures with common sigmoidal activation functions are compared, by deriving upper and lower bounds on their complexity, and studying how the complexity depends on the number of hidden units and the used activation function. The obtained results seem to support the idea that deep networks actually implements functions of higher complexity, so that they are able, with the same number of resources, to address more difficult problems.

Sentiment analysis: a comparison of deep learning neural network algorithm with SVM and naϊve Bayes for Indonesian text

NASA Astrophysics Data System (ADS)

Calvin Frans Mariel, Wahyu; Mariyah, Siti; Pramana, Setia

2018-03-01

Deep learning is a new era of machine learning techniques that essentially imitate the structure and function of the human brain. It is a development of deeper Artificial Neural Network (ANN) that uses more than one hidden layer. Deep Learning Neural Network has a great ability on recognizing patterns from various data types such as picture, audio, text, and many more. In this paper, the authors tries to measure that algorithm’s ability by applying it into the text classification. The classification task herein is done by considering the content of sentiment in a text which is also called as sentiment analysis. By using several combinations of text preprocessing and feature extraction techniques, we aim to compare the precise modelling results of Deep Learning Neural Network with the other two commonly used algorithms, the Naϊve Bayes and Support Vector Machine (SVM). This algorithm comparison uses Indonesian text data with balanced and unbalanced sentiment composition. Based on the experimental simulation, Deep Learning Neural Network clearly outperforms the Naϊve Bayes and SVM and offers a better F-1 Score while for the best feature extraction technique which improves that modelling result is Bigram.

A Grey Wolf Optimizer for Modular Granular Neural Networks for Human Recognition

PubMed Central

Sánchez, Daniela; Melin, Patricia

2017-01-01

A grey wolf optimizer for modular neural network (MNN) with a granular approach is proposed. The proposed method performs optimal granulation of data and design of modular neural networks architectures to perform human recognition, and to prove its effectiveness benchmark databases of ear, iris, and face biometric measures are used to perform tests and comparisons against other works. The design of a modular granular neural network (MGNN) consists in finding optimal parameters of its architecture; these parameters are the number of subgranules, percentage of data for the training phase, learning algorithm, goal error, number of hidden layers, and their number of neurons. Nowadays, there is a great variety of approaches and new techniques within the evolutionary computing area, and these approaches and techniques have emerged to help find optimal solutions to problems or models and bioinspired algorithms are part of this area. In this work a grey wolf optimizer is proposed for the design of modular granular neural networks, and the results are compared against a genetic algorithm and a firefly algorithm in order to know which of these techniques provides better results when applied to human recognition. PMID:28894461

A Grey Wolf Optimizer for Modular Granular Neural Networks for Human Recognition.

PubMed

Sánchez, Daniela; Melin, Patricia; Castillo, Oscar

2017-01-01

A grey wolf optimizer for modular neural network (MNN) with a granular approach is proposed. The proposed method performs optimal granulation of data and design of modular neural networks architectures to perform human recognition, and to prove its effectiveness benchmark databases of ear, iris, and face biometric measures are used to perform tests and comparisons against other works. The design of a modular granular neural network (MGNN) consists in finding optimal parameters of its architecture; these parameters are the number of subgranules, percentage of data for the training phase, learning algorithm, goal error, number of hidden layers, and their number of neurons. Nowadays, there is a great variety of approaches and new techniques within the evolutionary computing area, and these approaches and techniques have emerged to help find optimal solutions to problems or models and bioinspired algorithms are part of this area. In this work a grey wolf optimizer is proposed for the design of modular granular neural networks, and the results are compared against a genetic algorithm and a firefly algorithm in order to know which of these techniques provides better results when applied to human recognition.

Forecasting the portuguese stock market time series by using artificial neural networks

NASA Astrophysics Data System (ADS)

Isfan, Monica; Menezes, Rui; Mendes, Diana A.

2010-04-01

In this paper, we show that neural networks can be used to uncover the non-linearity that exists in the financial data. First, we follow a traditional approach by analysing the deterministic/stochastic characteristics of the Portuguese stock market data and some typical features are studied, like the Hurst exponents, among others. We also simulate a BDS test to investigate nonlinearities and the results are as expected: the financial time series do not exhibit linear dependence. Secondly, we trained four types of neural networks for the stock markets and used the models to make forecasts. The artificial neural networks were obtained using a three-layer feed-forward topology and the back-propagation learning algorithm. The quite large number of parameters that must be selected to develop a neural network forecasting model involves some trial and as a consequence the error is not small enough. In order to improve this we use a nonlinear optimization algorithm to minimize the error. Finally, the output of the 4 models is quite similar, leading to a qualitative forecast that we compare with the results of the application of k-nearest-neighbor for the same time series.

Convolutional neural networks for event-related potential detection: impact of the architecture.

PubMed

Cecotti, H

2017-07-01

The detection of brain responses at the single-trial level in the electroencephalogram (EEG) such as event-related potentials (ERPs) is a difficult problem that requires different processing steps to extract relevant discriminant features. While most of the signal and classification techniques for the detection of brain responses are based on linear algebra, different pattern recognition techniques such as convolutional neural network (CNN), as a type of deep learning technique, have shown some interests as they are able to process the signal after limited pre-processing. In this study, we propose to investigate the performance of CNNs in relation of their architecture and in relation to how they are evaluated: a single system for each subject, or a system for all the subjects. More particularly, we want to address the change of performance that can be observed between specifying a neural network to a subject, or by considering a neural network for a group of subjects, taking advantage of a larger number of trials from different subjects. The results support the conclusion that a convolutional neural network trained on different subjects can lead to an AUC above 0.9 by using an appropriate architecture using spatial filtering and shift invariant layers.

Exploring the Combination of Dempster-Shafer Theory and Neural Network for Predicting Trust and Distrust

PubMed Central

Wang, Xin; Wang, Ying; Sun, Hongbin

2016-01-01

In social media, trust and distrust among users are important factors in helping users make decisions, dissect information, and receive recommendations. However, the sparsity and imbalance of social relations bring great difficulties and challenges in predicting trust and distrust. Meanwhile, there are numerous inducing factors to determine trust and distrust relations. The relationship among inducing factors may be dependency, independence, and conflicting. Dempster-Shafer theory and neural network are effective and efficient strategies to deal with these difficulties and challenges. In this paper, we study trust and distrust prediction based on the combination of Dempster-Shafer theory and neural network. We firstly analyze the inducing factors about trust and distrust, namely, homophily, status theory, and emotion tendency. Then, we quantify inducing factors of trust and distrust, take these features as evidences, and construct evidence prototype as input nodes of multilayer neural network. Finally, we propose a framework of predicting trust and distrust which uses multilayer neural network to model the implementing process of Dempster-Shafer theory in different hidden layers, aiming to overcome the disadvantage of Dempster-Shafer theory without optimization method. Experimental results on a real-world dataset demonstrate the effectiveness of the proposed framework. PMID:27034651

Artificial neural networks for modeling time series of beach litter in the southern North Sea.

PubMed

Schulz, Marcus; Matthies, Michael

2014-07-01

In European marine waters, existing monitoring programs of beach litter need to be improved concerning litter items used as indicators of pollution levels, efficiency, and effectiveness. In order to ease and focus future monitoring of beach litter on few important litter items, feed-forward neural networks consisting of three layers were developed to relate single litter items to general categories of marine litter. The neural networks developed were applied to seven beaches in the southern North Sea and modeled time series of five general categories of marine litter, such as litter from fishing, shipping, and tourism. Results of regression analyses show that general categories were predicted significantly moderately to well. Measured and modeled data were in the same order of magnitude, and minima and maxima overlapped well. Neural networks were found to be eligible tools to deliver reliable predictions of marine litter with low computational effort and little input of information. Copyright © 2014 Elsevier Ltd. All rights reserved.

Human fatigue expression recognition through image-based dynamic multi-information and bimodal deep learning

NASA Astrophysics Data System (ADS)

Zhao, Lei; Wang, Zengcai; Wang, Xiaojin; Qi, Yazhou; Liu, Qing; Zhang, Guoxin

2016-09-01

Human fatigue is an important cause of traffic accidents. To improve the safety of transportation, we propose, in this paper, a framework for fatigue expression recognition using image-based facial dynamic multi-information and a bimodal deep neural network. First, the landmark of face region and the texture of eye region, which complement each other in fatigue expression recognition, are extracted from facial image sequences captured by a single camera. Then, two stacked autoencoder neural networks are trained for landmark and texture, respectively. Finally, the two trained neural networks are combined by learning a joint layer on top of them to construct a bimodal deep neural network. The model can be used to extract a unified representation that fuses landmark and texture modalities together and classify fatigue expressions accurately. The proposed system is tested on a human fatigue dataset obtained from an actual driving environment. The experimental results demonstrate that the proposed method performs stably and robustly, and that the average accuracy achieves 96.2%.

Predicting wettability behavior of fluorosilica coated metal surface using optimum neural network

NASA Astrophysics Data System (ADS)

Taghipour-Gorjikolaie, Mehran; Valipour Motlagh, Naser

2018-02-01

The interaction between variables, which are effective on the surface wettability, is very complex to predict the contact angles and sliding angles of liquid drops. In this paper, in order to solve this complexity, artificial neural network was used to develop reliable models for predicting the angles of liquid drops. Experimental data are divided into training data and testing data. By using training data and feed forward structure for the neural network and using particle swarm optimization for training the neural network based models, the optimum models were developed. The obtained results showed that regression index for the proposed models for the contact angles and sliding angles are 0.9874 and 0.9920, respectively. As it can be seen, these values are close to unit and it means the reliable performance of the models. Also, it can be inferred from the results that the proposed model have more reliable performance than multi-layer perceptron and radial basis function based models.

Functional model of biological neural networks.

PubMed

Lo, James Ting-Ho

2010-12-01

A functional model of biological neural networks, called temporal hierarchical probabilistic associative memory (THPAM), is proposed in this paper. THPAM comprises functional models of dendritic trees for encoding inputs to neurons, a first type of neuron for generating spike trains, a second type of neuron for generating graded signals to modulate neurons of the first type, supervised and unsupervised Hebbian learning mechanisms for easy learning and retrieving, an arrangement of dendritic trees for maximizing generalization, hardwiring for rotation-translation-scaling invariance, and feedback connections with different delay durations for neurons to make full use of present and past informations generated by neurons in the same and higher layers. These functional models and their processing operations have many functions of biological neural networks that have not been achieved by other models in the open literature and provide logically coherent answers to many long-standing neuroscientific questions. However, biological justifications of these functional models and their processing operations are required for THPAM to qualify as a macroscopic model (or low-order approximate) of biological neural networks.

Kannada character recognition system using neural network

NASA Astrophysics Data System (ADS)

Kumar, Suresh D. S.; Kamalapuram, Srinivasa K.; Kumar, Ajay B. R.

2013-03-01

Handwriting recognition has been one of the active and challenging research areas in the field of pattern recognition. It has numerous applications which include, reading aid for blind, bank cheques and conversion of any hand written document into structural text form. As there is no sufficient number of works on Indian language character recognition especially Kannada script among 15 major scripts in India. In this paper an attempt is made to recognize handwritten Kannada characters using Feed Forward neural networks. A handwritten Kannada character is resized into 20x30 Pixel. The resized character is used for training the neural network. Once the training process is completed the same character is given as input to the neural network with different set of neurons in hidden layer and their recognition accuracy rate for different Kannada characters has been calculated and compared. The results show that the proposed system yields good recognition accuracy rates comparable to that of other handwritten character recognition systems.

Adaptively combined FIR and functional link artificial neural network equalizer for nonlinear communication channel.

PubMed

Zhao, Haiquan; Zhang, Jiashu

2009-04-01

This paper proposes a novel computational efficient adaptive nonlinear equalizer based on combination of finite impulse response (FIR) filter and functional link artificial neural network (CFFLANN) to compensate linear and nonlinear distortions in nonlinear communication channel. This convex nonlinear combination results in improving the speed while retaining the lower steady-state error. In addition, since the CFFLANN needs not the hidden layers, which exist in conventional neural-network-based equalizers, it exhibits a simpler structure than the traditional neural networks (NNs) and can require less computational burden during the training mode. Moreover, appropriate adaptation algorithm for the proposed equalizer is derived by the modified least mean square (MLMS). Results obtained from the simulations clearly show that the proposed equalizer using the MLMS algorithm can availably eliminate various intensity linear and nonlinear distortions, and be provided with better anti-jamming performance. Furthermore, comparisons of the mean squared error (MSE), the bit error rate (BER), and the effect of eigenvalue ratio (EVR) of input correlation matrix are presented.

Influence of quality of images recorded in far infrared on pattern recognition based on neural networks and Eigenfaces algorithm

NASA Astrophysics Data System (ADS)

Jelen, Lukasz; Kobel, Joanna; Podbielska, Halina

2003-11-01

This paper discusses the possibility of exploiting of the tennovision registration and artificial neural networks for facial recognition systems. A biometric system that is able to identify people from thermograms is presented. To identify a person we used the Eigenfaces algorithm. For the face detection in the picture the backpropagation neural network was designed. For this purpose thermograms of 10 people in various external conditions were studies. The Eigenfaces algorithm calculated an average face and then the set of characteristic features for each studied person was produced. The neural network has to detect the face in the image before it actually can be identified. We used five hidden layers for that purpose. It was shown that the errors in recognition depend on the feature extraction, for low quality pictures the error was so high as 30%. However, for pictures with a good feature extraction the results of proper identification higher then 90%, were obtained.

Beyond Retinal Layers: A Deep Voting Model for Automated Geographic Atrophy Segmentation in SD-OCT Images

PubMed Central

Ji, Zexuan; Chen, Qiang; Niu, Sijie; Leng, Theodore; Rubin, Daniel L.

2018-01-01

Purpose To automatically and accurately segment geographic atrophy (GA) in spectral-domain optical coherence tomography (SD-OCT) images by constructing a voting system with deep neural networks without the use of retinal layer segmentation. Methods An automatic GA segmentation method for SD-OCT images based on the deep network was constructed. The structure of the deep network was composed of five layers, including one input layer, three hidden layers, and one output layer. During the training phase, the labeled A-scans with 1024 features were directly fed into the network as the input layer to obtain the deep representations. Then a soft-max classifier was trained to determine the label of each individual pixel. Finally, a voting decision strategy was used to refine the segmentation results among 10 trained models. Results Two image data sets with GA were used to evaluate the model. For the first dataset, our algorithm obtained a mean overlap ratio (OR) 86.94% ± 8.75%, absolute area difference (AAD) 11.49% ± 11.50%, and correlation coefficients (CC) 0.9857; for the second dataset, the mean OR, AAD, and CC of the proposed method were 81.66% ± 10.93%, 8.30% ± 9.09%, and 0.9952, respectively. The proposed algorithm was capable of improving over 5% and 10% segmentation accuracy, respectively, when compared with several state-of-the-art algorithms on two data sets. Conclusions Without retinal layer segmentation, the proposed algorithm could produce higher segmentation accuracy and was more stable when compared with state-of-the-art methods that relied on retinal layer segmentation results. Our model may provide reliable GA segmentations from SD-OCT images and be useful in the clinical diagnosis of advanced nonexudative AMD. Translational Relevance Based on the deep neural networks, this study presents an accurate GA segmentation method for SD-OCT images without using any retinal layer segmentation results, and may contribute to improved understanding of advanced nonexudative AMD. PMID:29302382

Beyond Retinal Layers: A Deep Voting Model for Automated Geographic Atrophy Segmentation in SD-OCT Images.

PubMed

Ji, Zexuan; Chen, Qiang; Niu, Sijie; Leng, Theodore; Rubin, Daniel L

2018-01-01

To automatically and accurately segment geographic atrophy (GA) in spectral-domain optical coherence tomography (SD-OCT) images by constructing a voting system with deep neural networks without the use of retinal layer segmentation. An automatic GA segmentation method for SD-OCT images based on the deep network was constructed. The structure of the deep network was composed of five layers, including one input layer, three hidden layers, and one output layer. During the training phase, the labeled A-scans with 1024 features were directly fed into the network as the input layer to obtain the deep representations. Then a soft-max classifier was trained to determine the label of each individual pixel. Finally, a voting decision strategy was used to refine the segmentation results among 10 trained models. Two image data sets with GA were used to evaluate the model. For the first dataset, our algorithm obtained a mean overlap ratio (OR) 86.94% ± 8.75%, absolute area difference (AAD) 11.49% ± 11.50%, and correlation coefficients (CC) 0.9857; for the second dataset, the mean OR, AAD, and CC of the proposed method were 81.66% ± 10.93%, 8.30% ± 9.09%, and 0.9952, respectively. The proposed algorithm was capable of improving over 5% and 10% segmentation accuracy, respectively, when compared with several state-of-the-art algorithms on two data sets. Without retinal layer segmentation, the proposed algorithm could produce higher segmentation accuracy and was more stable when compared with state-of-the-art methods that relied on retinal layer segmentation results. Our model may provide reliable GA segmentations from SD-OCT images and be useful in the clinical diagnosis of advanced nonexudative AMD. Based on the deep neural networks, this study presents an accurate GA segmentation method for SD-OCT images without using any retinal layer segmentation results, and may contribute to improved understanding of advanced nonexudative AMD.

Artificial Neural Networks: A Novel Approach to Analysing the Nutritional Ecology of a Blowfly Species, Chrysomya megacephala

PubMed Central

Bianconi, André; Zuben, Cláudio J. Von; Serapião, Adriane B. de S.; Govone, José S.

2010-01-01

Bionomic features of blowflies may be clarified and detailed by the deployment of appropriate modelling techniques such as artificial neural networks, which are mathematical tools widely applied to the resolution of complex biological problems. The principal aim of this work was to use three well-known neural networks, namely Multi-Layer Perceptron (MLP), Radial Basis Function (RBF), and Adaptive Neural Network-Based Fuzzy Inference System (ANFIS), to ascertain whether these tools would be able to outperform a classical statistical method (multiple linear regression) in the prediction of the number of resultant adults (survivors) of experimental populations of Chrysomya megacephala (F.) (Diptera: Calliphoridae), based on initial larval density (number of larvae), amount of available food, and duration of immature stages. The coefficient of determination (R2) derived from the RBF was the lowest in the testing subset in relation to the other neural networks, even though its R2 in the training subset exhibited virtually a maximum value. The ANFIS model permitted the achievement of the best testing performance. Hence this model was deemed to be more effective in relation to MLP and RBF for predicting the number of survivors. All three networks outperformed the multiple linear regression, indicating that neural models could be taken as feasible techniques for predicting bionomic variables concerning the nutritional dynamics of blowflies. PMID:20569135

Dense modifiable interconnections utilizing photorefractive volume holograms

NASA Astrophysics Data System (ADS)

Psaltis, Demetri; Qiao, Yong

1990-11-01

This report describes an experimental two-layer optical neural network built at Caltech. The system uses photorefractive volume holograms to implement dense, modifiable synaptic interconnections and liquid crystal light valves (LCVS) to perform nonlinear thresholding operations. Kanerva's Sparse, Distributed Memory was implemented using this network and its ability to recognize handwritten character-alphabet (A-Z) has been demonstrated experimentally. According to Kanerva's model, the first layer has fixed, random weights of interconnections and the second layer is trained by sum-of-outer-products rule. After training, the recognition rates of the network on the training set (104 patterns) and test set (520 patterns) are 100 and 50 percent, respectively.

Laser programmable integrated circuit for forming synapses in neural networks

DOEpatents

Fu, C.Y.

1997-02-11

Customizable neural network in which one or more resistors form each synapse is disclosed. All the resistors in the synaptic array are identical, thus simplifying the processing issues. Highly doped, amorphous silicon is used as the resistor material, to create extremely high resistances occupying very small spaces. Connected in series with each resistor in the array is at least one severable conductor whose uppermost layer has a lower reflectivity of laser energy than typical metal conductors at a desired laser wavelength. 5 figs.

Process for forming synapses in neural networks and resistor therefor

DOEpatents

Fu, C.Y.

1996-07-23

Customizable neural network in which one or more resistors form each synapse is disclosed. All the resistors in the synaptic array are identical, thus simplifying the processing issues. Highly doped, amorphous silicon is used as the resistor material, to create extremely high resistances occupying very small spaces. Connected in series with each resistor in the array is at least one severable conductor whose uppermost layer has a lower reflectivity of laser energy than typical metal conductors at a desired laser wavelength. 5 figs.

Laser programmable integrated curcuit for forming synapses in neural networks

DOEpatents

Fu, Chi Y.

1997-01-01

Customizable neural network in which one or more resistors form each synapse. All the resistors in the synaptic array are identical, thus simplifying the processing issues. Highly doped, amorphous silicon is used as the resistor material, to create extremely high resistances occupying very small spaces. Connected in series with each resistor in the array is at least one severable conductor whose uppermost layer has a lower reflectivity of laser energy than typical metal conductors at a desired laser wavelength.

Process for forming synapses in neural networks and resistor therefor

DOEpatents

Fu, Chi Y.

1996-01-01

Customizable neural network in which one or more resistors form each synapse. All the resistors in the synaptic array are identical, thus simplifying the processing issues. Highly doped, amorphous silicon is used as the resistor material, to create extremely high resistances occupying very small spaces. Connected in series with each resistor in the array is at least one severable conductor whose uppermost layer has a lower reflectivity of laser energy than typical metal conductors at a desired laser wavelength.

Modelization of three-layered polymer coated steel-strip ironing process using a neural network

NASA Astrophysics Data System (ADS)

Sellés, M. A.; Schmid, S. R.; Sánchez-Caballero, S.; Seguí, V. J.; Reig, M. J.; Pla, R.

2012-04-01

An alternative to the traditional can manufacturing process is to use plastic laminated rolled steels as base stocks. This material consist of pre-heated steel coils that are sandwiched between one or two sheets of polymer. The heated sheets are then immediately quenched, which yields a strong bond between the layers. Such polymer-coated steels were investigated by Jaworski [1,2] and Sellés [3], and found to be suitable for ironing with carefully controlled conditions. A novel multi-layer polymer coated steel has been developed for container applications. This material presents an interesting extension to previous research on polymer laminated steel in ironing, and offers several advantages over the previous material (Sellés [3]). This document shows a modelization for the ironing process (the most crucial step in can manufacturing) done by using a neural network

Parsing recursive sentences with a connectionist model including a neural stack and synaptic gating.

PubMed

Fedor, Anna; Ittzés, Péter; Szathmáry, Eörs

2011-02-21

It is supposed that humans are genetically predisposed to be able to recognize sequences of context-free grammars with centre-embedded recursion while other primates are restricted to the recognition of finite state grammars with tail-recursion. Our aim was to construct a minimalist neural network that is able to parse artificial sentences of both grammars in an efficient way without using the biologically unrealistic backpropagation algorithm. The core of this network is a neural stack-like memory where the push and pop operations are regulated by synaptic gating on the connections between the layers of the stack. The network correctly categorizes novel sentences of both grammars after training. We suggest that the introduction of the neural stack memory will turn out to be substantial for any biological 'hierarchical processor' and the minimalist design of the model suggests a quest for similar, realistic neural architectures. Copyright © 2010 Elsevier Ltd. All rights reserved.

Learning a single-hidden layer feedforward neural network using a rank correlation-based strategy with application to high dimensional gene expression and proteomic spectra datasets in cancer detection.

PubMed

Belciug, Smaranda; Gorunescu, Florin

2018-06-08

Methods based on microarrays (MA), mass spectrometry (MS), and machine learning (ML) algorithms have evolved rapidly in recent years, allowing for early detection of several types of cancer. A pitfall of these approaches, however, is the overfitting of data due to large number of attributes and small number of instances -- a phenomenon known as the 'curse of dimensionality'. A potentially fruitful idea to avoid this drawback is to develop algorithms that combine fast computation with a filtering module for the attributes. The goal of this paper is to propose a statistical strategy to initiate the hidden nodes of a single-hidden layer feedforward neural network (SLFN) by using both the knowledge embedded in data and a filtering mechanism for attribute relevance. In order to attest its feasibility, the proposed model has been tested on five publicly available high-dimensional datasets: breast, lung, colon, and ovarian cancer regarding gene expression and proteomic spectra provided by cDNA arrays, DNA microarray, and MS. The novel algorithm, called adaptive SLFN (aSLFN), has been compared with four major classification algorithms: traditional ELM, radial basis function network (RBF), single-hidden layer feedforward neural network trained by backpropagation algorithm (BP-SLFN), and support vector-machine (SVM). Experimental results showed that the classification performance of aSLFN is competitive with the comparison models. Copyright © 2018. Published by Elsevier Inc.

Use of a Neural Net to Model the Impact of Optical Coherence Tomography Abnormalities on Vision in Age-related Macular Degeneration.

PubMed

Aslam, Tariq M; Zaki, Haider R; Mahmood, Sajjad; Ali, Zaria C; Ahmad, Nur A; Thorell, Mariana R; Balaskas, Konstantinos

2018-01-01

To develop a neural network for the estimation of visual acuity from optical coherence tomography (OCT) images of patients with neovascular age-related macular degeneration (AMD) and to demonstrate its use to model the impact of specific controlled OCT changes on vision. Artificial intelligence (neural network) study. We assessed 1400 OCT scans of patients with neovascular AMD. Fifteen physical features for each eligible OCT, as well as patient age, were used as input data and corresponding recorded visual acuity as the target data to train, validate, and test a supervised neural network. We then applied this network to model the impact on acuity of defined OCT changes in subretinal fluid, subretinal hyperreflective material, and loss of external limiting membrane (ELM) integrity. A total of 1210 eligible OCT scans were analyzed, resulting in 1210 data points, which were each 16-dimensional. A 10-layer feed-forward neural network with 1 hidden layer of 10 neurons was trained to predict acuity and demonstrated a root mean square error of 8.2 letters for predicted compared to actual visual acuity and a mean regression coefficient of 0.85. A virtual model using this network demonstrated the relationship of visual acuity to specific, programmed changes in OCT characteristics. When ELM is intact, there is a shallow decline in acuity with increasing subretinal fluid but a much steeper decline with equivalent increasing subretinal hyperreflective material. When ELM is not intact, all visual acuities are reduced. Increasing subretinal hyperreflective material or subretinal fluid in this circumstance reduces vision further still, but with a smaller gradient than when ELM is intact. The supervised machine learning neural network developed is able to generate an estimated visual acuity value from OCT images in a population of patients with AMD. These findings should be of clinical and research interest in macular degeneration, for example in estimating visual prognosis or highlighting the importance of developing treatments targeting more visually destructive pathologies. Copyright © 2017 Elsevier Inc. All rights reserved.

Spiking neural network model for memorizing sequences with forward and backward recall.

PubMed

Borisyuk, Roman; Chik, David; Kazanovich, Yakov; da Silva Gomes, João

2013-06-01

We present an oscillatory network of conductance based spiking neurons of Hodgkin-Huxley type as a model of memory storage and retrieval of sequences of events (or objects). The model is inspired by psychological and neurobiological evidence on sequential memories. The building block of the model is an oscillatory module which contains excitatory and inhibitory neurons with all-to-all connections. The connection architecture comprises two layers. A lower layer represents consecutive events during their storage and recall. This layer is composed of oscillatory modules. Plastic excitatory connections between the modules are implemented using an STDP type learning rule for sequential storage. Excitatory neurons in the upper layer project star-like modifiable connections toward the excitatory lower layer neurons. These neurons in the upper layer are used to tag sequences of events represented in the lower layer. Computer simulations demonstrate good performance of the model including difficult cases when different sequences contain overlapping events. We show that the model with STDP type or anti-STDP type learning rules can be applied for the simulation of forward and backward replay of neural spikes respectively. Copyright © 2013 Elsevier Ireland Ltd. All rights reserved.

Understanding auditory distance estimation by humpback whales: a computational approach.

PubMed

Mercado, E; Green, S R; Schneider, J N

2008-02-01

Ranging, the ability to judge the distance to a sound source, depends on the presence of predictable patterns of attenuation. We measured long-range sound propagation in coastal waters to assess whether humpback whales might use frequency degradation cues to range singing whales. Two types of neural networks, a multi-layer and a single-layer perceptron, were trained to classify recorded sounds by distance traveled based on their frequency content. The multi-layer network successfully classified received sounds, demonstrating that the distorting effects of underwater propagation on frequency content provide sufficient cues to estimate source distance. Normalizing received sounds with respect to ambient noise levels increased the accuracy of distance estimates by single-layer perceptrons, indicating that familiarity with background noise can potentially improve a listening whale's ability to range. To assess whether frequency patterns predictive of source distance were likely to be perceived by whales, recordings were pre-processed using a computational model of the humpback whale's peripheral auditory system. Although signals processed with this model contained less information than the original recordings, neural networks trained with these physiologically based representations estimated source distance more accurately, suggesting that listening whales should be able to range singers using distance-dependent changes in frequency content.

Neural Networks and other Techniques for Fault Identification and Isolation of Aircraft Systems

NASA Technical Reports Server (NTRS)

Innocenti, M.; Napolitano, M.

2003-01-01

Fault identification, isolation, and accomodation have become critical issues in the overall performance of advanced aircraft systems. Neural Networks have shown to be a very attractive alternative to classic adaptation methods for identification and control of non-linear dynamic systems. The purpose of this paper is to show the improvements in neural network applications achievable through the use of learning algorithms more efficient than the classic Back-Propagation, and through the implementation of the neural schemes in parallel hardware. The results of the analysis of a scheme for Sensor Failure, Detection, Identification and Accommodation (SFDIA) using experimental flight data of a research aircraft model are presented. Conventional approaches to the problem are based on observers and Kalman Filters while more recent methods are based on neural approximators. The work described in this paper is based on the use of neural networks (NNs) as on-line learning non-linear approximators. The performances of two different neural architectures were compared. The first architecture is based on a Multi Layer Perceptron (MLP) NN trained with the Extended Back Propagation algorithm (EBPA). The second architecture is based on a Radial Basis Function (RBF) NN trained with the Extended-MRAN (EMRAN) algorithms. In addition, alternative methods for communications links fault detection and accomodation are presented, relative to multiple unmanned aircraft applications.

A neural network device for on-line particle identification in cosmic ray experiments

NASA Astrophysics Data System (ADS)

Scrimaglio, R.; Finetti, N.; D'Altorio, L.; Rantucci, E.; Raso, M.; Segreto, E.; Tassoni, A.; Cardarilli, G. C.

2004-05-01

On-line particle identification is one of the main goals of many experiments in space both for rare event studies and for optimizing measurements along the orbital trajectory. Neural networks can be a useful tool for signal processing and real time data analysis in such experiments. In this document we report on the performances of a programmable neural device which was developed in VLSI analog/digital technology. Neurons and synapses were accomplished by making use of Operational Transconductance Amplifier (OTA) structures. In this paper we report on the results of measurements performed in order to verify the agreement of the characteristic curves of each elementary cell with simulations and on the device performances obtained by implementing simple neural structures on the VLSI chip. A feed-forward neural network (Multi-Layer Perceptron, MLP) was implemented on the VLSI chip and trained to identify particles by processing the signals of two-dimensional position-sensitive Si detectors. The radiation monitoring device consisted of three double-sided silicon strip detectors. From the analysis of a set of simulated data it was found that the MLP implemented on the neural device gave results comparable with those obtained with the standard method of analysis confirming that the implemented neural network could be employed for real time particle identification.

Artificial neural network modelling of uncertainty in gamma-ray spectrometry

NASA Astrophysics Data System (ADS)

Dragović, S.; Onjia, A.; Stanković, S.; Aničin, I.; Bačić, G.

2005-03-01

An artificial neural network (ANN) model for the prediction of measuring uncertainties in gamma-ray spectrometry was developed and optimized. A three-layer feed-forward ANN with back-propagation learning algorithm was used to model uncertainties of measurement of activity levels of eight radionuclides ( 226Ra, 238U, 235U, 40K, 232Th, 134Cs, 137Cs and 7Be) in soil samples as a function of measurement time. It was shown that the neural network provides useful data even from small experimental databases. The performance of the optimized neural network was found to be very good, with correlation coefficients ( R2) between measured and predicted uncertainties ranging from 0.9050 to 0.9915. The correlation coefficients did not significantly deteriorate when the network was tested on samples with greatly different uranium-to-thorium ( 238U/ 232Th) ratios. The differences between measured and predicted uncertainties were not influenced by the absolute values of uncertainties of measured radionuclide activities. Once the ANN is trained, it could be employed in analyzing soil samples regardless of the 238U/ 232Th ratio. It was concluded that a considerable saving in time could be obtained using the trained neural network model for predicting the measurement times needed to attain the desired statistical accuracy.

Neural network method for lossless two-conductor transmission line equations based on the IELM algorithm

NASA Astrophysics Data System (ADS)

Yang, Yunlei; Hou, Muzhou; Luo, Jianshu; Liu, Taohua

2018-06-01

With the increasing demands for vast amounts of data and high-speed signal transmission, the use of multi-conductor transmission lines is becoming more common. The impact of transmission lines on signal transmission is thus a key issue affecting the performance of high-speed digital systems. To solve the problem of lossless two-conductor transmission line equations (LTTLEs), a neural network model and algorithm are explored in this paper. By selecting the product of two triangular basis functions as the activation function of hidden layer neurons, we can guarantee the separation of time, space, and phase orthogonality. By adding the initial condition to the neural network, an improved extreme learning machine (IELM) algorithm for solving the network weight is obtained. This is different to the traditional method for converting the initial condition into the iterative constraint condition. Calculation software for solving the LTTLEs based on the IELM algorithm is developed. Numerical experiments show that the results are consistent with those of the traditional method. The proposed neural network algorithm can find the terminal voltage of the transmission line and also the voltage of any observation point. It is possible to calculate the value at any given point by using the neural network model to solve the transmission line equation.

Neural network simulation of the atmospheric point spread function for the adjacency effect research

NASA Astrophysics Data System (ADS)

Ma, Xiaoshan; Wang, Haidong; Li, Ligang; Yang, Zhen; Meng, Xin

2016-10-01

Adjacency effect could be regarded as the convolution of the atmospheric point spread function (PSF) and the surface leaving radiance. Monte Carlo is a common method to simulate the atmospheric PSF. But it can't obtain analytic expression and the meaningful results can be only acquired by statistical analysis of millions of data. A backward Monte Carlo algorithm was employed to simulate photon emitting and propagating in the atmosphere under different conditions. The PSF was determined by recording the photon-receiving numbers in fixed bin at different position. A multilayer feed-forward neural network with a single hidden layer was designed to learn the relationship between the PSF's and the input condition parameters. The neural network used the back-propagation learning rule for training. Its input parameters involved atmosphere condition, spectrum range, observing geometry. The outputs of the network were photon-receiving numbers in the corresponding bin. Because the output units were too many to be allowed by neural network, the large network was divided into a collection of smaller ones. These small networks could be ran simultaneously on many workstations and/or PCs to speed up the training. It is important to note that the simulated PSF's by Monte Carlo technique in non-nadir viewing angles are more complicated than that in nadir conditions which brings difficulties in the design of the neural network. The results obtained show that the neural network approach could be very useful to compute the atmospheric PSF based on the simulated data generated by Monte Carlo method.

Sensitivity Analysis for Probabilistic Neural Network Structure Reduction.

PubMed

Kowalski, Piotr A; Kusy, Maciej

2018-05-01

In this paper, we propose the use of local sensitivity analysis (LSA) for the structure simplification of the probabilistic neural network (PNN). Three algorithms are introduced. The first algorithm applies LSA to the PNN input layer reduction by selecting significant features of input patterns. The second algorithm utilizes LSA to remove redundant pattern neurons of the network. The third algorithm combines the proposed two and constitutes the solution of how they can work together. PNN with a product kernel estimator is used, where each multiplicand computes a one-dimensional Cauchy function. Therefore, the smoothing parameter is separately calculated for each dimension by means of the plug-in method. The classification qualities of the reduced and full structure PNN are compared. Furthermore, we evaluate the performance of PNN, for which global sensitivity analysis (GSA) and the common reduction methods are applied, both in the input layer and the pattern layer. The models are tested on the classification problems of eight repository data sets. A 10-fold cross validation procedure is used to determine the prediction ability of the networks. Based on the obtained results, it is shown that the LSA can be used as an alternative PNN reduction approach.

Electron beam lithographic modeling assisted by artificial intelligence technology

NASA Astrophysics Data System (ADS)

Nakayamada, Noriaki; Nishimura, Rieko; Miura, Satoru; Nomura, Haruyuki; Kamikubo, Takashi

2017-07-01

We propose a new concept of tuning a point-spread function (a "kernel" function) in the modeling of electron beam lithography using the machine learning scheme. Normally in the work of artificial intelligence, the researchers focus on the output results from a neural network, such as success ratio in image recognition or improved production yield, etc. In this work, we put more focus on the weights connecting the nodes in a convolutional neural network, which are naturally the fractions of a point-spread function, and take out those weighted fractions after learning to be utilized as a tuned kernel. Proof-of-concept of the kernel tuning has been demonstrated using the examples of proximity effect correction with 2-layer network, and charging effect correction with 3-layer network. This type of new tuning method can be beneficial to give researchers more insights to come up with a better model, yet it might be too early to be deployed to production to give better critical dimension (CD) and positional accuracy almost instantly.

DeepFix: A Fully Convolutional Neural Network for Predicting Human Eye Fixations.

PubMed

Kruthiventi, Srinivas S S; Ayush, Kumar; Babu, R Venkatesh

2017-09-01

Understanding and predicting the human visual attention mechanism is an active area of research in the fields of neuroscience and computer vision. In this paper, we propose DeepFix, a fully convolutional neural network, which models the bottom-up mechanism of visual attention via saliency prediction. Unlike classical works, which characterize the saliency map using various hand-crafted features, our model automatically learns features in a hierarchical fashion and predicts the saliency map in an end-to-end manner. DeepFix is designed to capture semantics at multiple scales while taking global context into account, by using network layers with very large receptive fields. Generally, fully convolutional nets are spatially invariant-this prevents them from modeling location-dependent patterns (e.g., centre-bias). Our network handles this by incorporating a novel location-biased convolutional layer. We evaluate our model on multiple challenging saliency data sets and show that it achieves the state-of-the-art results.

Image inpainting and super-resolution using non-local recursive deep convolutional network with skip connections

NASA Astrophysics Data System (ADS)

Liu, Miaofeng

2017-07-01

In recent years, deep convolutional neural networks come into use in image inpainting and super-resolution in many fields. Distinct to most of the former methods requiring to know beforehand the local information for corrupted pixels, we propose a 20-depth fully convolutional network to learn an end-to-end mapping a dataset of damaged/ground truth subimage pairs realizing non-local blind inpainting and super-resolution. As there often exist image with huge corruptions or inpainting on a low-resolution image that the existing approaches unable to perform well, we also share parameters in local area of layers to achieve spatial recursion and enlarge the receptive field. To avoid the difficulty of training this deep neural network, skip-connections between symmetric convolutional layers are designed. Experimental results shows that the proposed method outperforms state-of-the-art methods for diverse corrupting and low-resolution conditions, it works excellently when realizing super-resolution and image inpainting simultaneously

What the success of brain imaging implies about the neural code

PubMed Central

Guest, Olivia; Love, Bradley C

2017-01-01

The success of fMRI places constraints on the nature of the neural code. The fact that researchers can infer similarities between neural representations, despite fMRI’s limitations, implies that certain neural coding schemes are more likely than others. For fMRI to succeed given its low temporal and spatial resolution, the neural code must be smooth at the voxel and functional level such that similar stimuli engender similar internal representations. Through proof and simulation, we determine which coding schemes are plausible given both fMRI’s successes and its limitations in measuring neural activity. Deep neural network approaches, which have been forwarded as computational accounts of the ventral stream, are consistent with the success of fMRI, though functional smoothness breaks down in the later network layers. These results have implications for the nature of the neural code and ventral stream, as well as what can be successfully investigated with fMRI. DOI: http://dx.doi.org/10.7554/eLife.21397.001 PMID:28103186

Neural networks with multiple general neuron models: a hybrid computational intelligence approach using Genetic Programming.

PubMed

Barton, Alan J; Valdés, Julio J; Orchard, Robert

2009-01-01

Classical neural networks are composed of neurons whose nature is determined by a certain function (the neuron model), usually pre-specified. In this paper, a type of neural network (NN-GP) is presented in which: (i) each neuron may have its own neuron model in the form of a general function, (ii) any layout (i.e network interconnection) is possible, and (iii) no bias nodes or weights are associated to the connections, neurons or layers. The general functions associated to a neuron are learned by searching a function space. They are not provided a priori, but are rather built as part of an Evolutionary Computation process based on Genetic Programming. The resulting network solutions are evaluated based on a fitness measure, which may, for example, be based on classification or regression errors. Two real-world examples are presented to illustrate the promising behaviour on classification problems via construction of a low-dimensional representation of a high-dimensional parameter space associated to the set of all network solutions.

Neural network for nonsmooth pseudoconvex optimization with general convex constraints.

PubMed

Bian, Wei; Ma, Litao; Qin, Sitian; Xue, Xiaoping

2018-05-01

In this paper, a one-layer recurrent neural network is proposed for solving a class of nonsmooth, pseudoconvex optimization problems with general convex constraints. Based on the smoothing method, we construct a new regularization function, which does not depend on any information of the feasible region. Thanks to the special structure of the regularization function, we prove the global existence, uniqueness and "slow solution" character of the state of the proposed neural network. Moreover, the state solution of the proposed network is proved to be convergent to the feasible region in finite time and to the optimal solution set of the related optimization problem subsequently. In particular, the convergence of the state to an exact optimal solution is also considered in this paper. Numerical examples with simulation results are given to show the efficiency and good characteristics of the proposed network. In addition, some preliminary theoretical analysis and application of the proposed network for a wider class of dynamic portfolio optimization are included. Copyright © 2018 Elsevier Ltd. All rights reserved.

Implementing a Bayes Filter in a Neural Circuit: The Case of Unknown Stimulus Dynamics.

PubMed

Sokoloski, Sacha

2017-09-01

In order to interact intelligently with objects in the world, animals must first transform neural population responses into estimates of the dynamic, unknown stimuli that caused them. The Bayesian solution to this problem is known as a Bayes filter, which applies Bayes' rule to combine population responses with the predictions of an internal model. The internal model of the Bayes filter is based on the true stimulus dynamics, and in this note, we present a method for training a theoretical neural circuit to approximately implement a Bayes filter when the stimulus dynamics are unknown. To do this we use the inferential properties of linear probabilistic population codes to compute Bayes' rule and train a neural network to compute approximate predictions by the method of maximum likelihood. In particular, we perform stochastic gradient descent on the negative log-likelihood of the neural network parameters with a novel approximation of the gradient. We demonstrate our methods on a finite-state, a linear, and a nonlinear filtering problem and show how the hidden layer of the neural network develops tuning curves consistent with findings in experimental neuroscience.

Discrimination of liver cancer in cellular level based on backscatter micro-spectrum with PCA algorithm and BP neural network

NASA Astrophysics Data System (ADS)

Yang, Jing; Wang, Cheng; Cai, Gan; Dong, Xiaona

2016-10-01

The incidence and mortality rate of the primary liver cancer are very high and its postoperative metastasis and recurrence have become important factors to the prognosis of patients. Circulating tumor cells (CTC), as a new tumor marker, play important roles in the early diagnosis and individualized treatment. This paper presents an effective method to distinguish liver cancer based on the cellular scattering spectrum, which is a non-fluorescence technique based on the fiber confocal microscopic spectrometer. Combining the principal component analysis (PCA) with back propagation (BP) neural network were utilized to establish an automatic recognition model for backscatter spectrum of the liver cancer cells from blood cell. PCA was applied to reduce the dimension of the scattering spectral data which obtained by the fiber confocal microscopic spectrometer. After dimensionality reduction by PCA, a neural network pattern recognition model with 2 input layer nodes, 11 hidden layer nodes, 3 output nodes was established. We trained the network with 66 samples and also tested it. Results showed that the recognition rate of the three types of cells is more than 90%, the relative standard deviation is only 2.36%. The experimental results showed that the fiber confocal microscopic spectrometer combining with the algorithm of PCA and BP neural network can automatically identify the liver cancer cell from the blood cells. This will provide a better tool for investigating the metastasis of liver cancers in vivo, the biology metabolic characteristics of liver cancers and drug transportation. Additionally, it is obviously referential in practical application.

Radio Galaxy Zoo: compact and extended radio source classification with deep learning

NASA Astrophysics Data System (ADS)

Lukic, V.; Brüggen, M.; Banfield, J. K.; Wong, O. I.; Rudnick, L.; Norris, R. P.; Simmons, B.

2018-05-01

Machine learning techniques have been increasingly useful in astronomical applications over the last few years, for example in the morphological classification of galaxies. Convolutional neural networks have proven to be highly effective in classifying objects in image data. In the context of radio-interferometric imaging in astronomy, we looked for ways to identify multiple components of individual sources. To this effect, we design a convolutional neural network to differentiate between different morphology classes using sources from the Radio Galaxy Zoo (RGZ) citizen science project. In this first step, we focus on exploring the factors that affect the performance of such neural networks, such as the amount of training data, number and nature of layers, and the hyperparameters. We begin with a simple experiment in which we only differentiate between two extreme morphologies, using compact and multiple-component extended sources. We found that a three-convolutional layer architecture yielded very good results, achieving a classification accuracy of 97.4 per cent on a test data set. The same architecture was then tested on a four-class problem where we let the network classify sources into compact and three classes of extended sources, achieving a test accuracy of 93.5 per cent. The best-performing convolutional neural network set-up has been verified against RGZ Data Release 1 where a final test accuracy of 94.8 per cent was obtained, using both original and augmented images. The use of sigma clipping does not offer a significant benefit overall, except in cases with a small number of training images.

High pressure air compressor valve fault diagnosis using feedforward neural networks

NASA Astrophysics Data System (ADS)

James Li, C.; Yu, Xueli

1995-09-01

Feedforward neural networks (FNNs) are developed and implemented to classify a four-stage high pressure air compressor into one of the following conditions: baseline, suction or exhaust valve faults. These FNNs are used for the compressor's automatic condition monitoring and fault diagnosis. Measurements of 39 variables are obtained under different baseline conditions and third-stage suction and exhaust valve faults. These variables include pressures and temperatures at all stages, voltage between phase aand phase b, voltage between phase band phase c, total three-phase real power, cooling water flow rate, etc. To reduce the number of variables, the amount of their discriminatory information is quantified by scattering matrices to identify statistical significant ones. Measurements of the selected variables are then used by a fully automatic structural and weight learning algorithm to construct three-layer FNNs to classify the compressor's condition. This learning algorithm requires neither guesses of initial weight values nor number of neurons in the hidden layer of an FNN. It takes an incremental approach in which a hidden neuron is trained by exemplars and then augmented to the existing network. These exemplars are then made orthogonal to the newly identified hidden neuron. They are subsequently used for the training of the next hidden neuron. The betterment continues until a desired accuracy is reached. After the neural networks are established, novel measurements from various conditions that haven't been previously seen by the FNNs are then used to evaluate their ability in fault diagnosis. The trained neural networks provide very accurate diagnosis for suction and discharge valve defects.

A Neural Network/Acoustic Emission Analysis of Impact Damaged Graphite/Epoxy Pressure Vessels

NASA Technical Reports Server (NTRS)

Walker, James L.; Hill, Erik v. K.; Workman, Gary L.; Russell, Samuel S.

1995-01-01

Acoustic emission (AE) signal analysis has been used to measure the effects of impact damage on burst pressure in 5.75 inch diameter, inert propellant filled, filament wound pressure vessels. The AE data were collected from fifteen graphite/epoxy pressure vessels featuring five damage states and three resin systems. A burst pressure prediction model was developed by correlating the AE amplitude (frequency) distribution, generated during the first pressure ramp to 800 psig (approximately 25% of the average expected burst pressure for an undamaged vessel) to known burst pressures using a four layered back propagation neural network. The neural network, trained on three vessels from each resin system, was able to predict burst pressures with a worst case error of 5.7% for the entire fifteen bottle set.

A link prediction method for heterogeneous networks based on BP neural network

NASA Astrophysics Data System (ADS)

Li, Ji-chao; Zhao, Dan-ling; Ge, Bing-Feng; Yang, Ke-Wei; Chen, Ying-Wu

2018-04-01

Most real-world systems, composed of different types of objects connected via many interconnections, can be abstracted as various complex heterogeneous networks. Link prediction for heterogeneous networks is of great significance for mining missing links and reconfiguring networks according to observed information, with considerable applications in, for example, friend and location recommendations and disease-gene candidate detection. In this paper, we put forward a novel integrated framework, called MPBP (Meta-Path feature-based BP neural network model), to predict multiple types of links for heterogeneous networks. More specifically, the concept of meta-path is introduced, followed by the extraction of meta-path features for heterogeneous networks. Next, based on the extracted meta-path features, a supervised link prediction model is built with a three-layer BP neural network. Then, the solution algorithm of the proposed link prediction model is put forward to obtain predicted results by iteratively training the network. Last, numerical experiments on the dataset of examples of a gene-disease network and a combat network are conducted to verify the effectiveness and feasibility of the proposed MPBP. It shows that the MPBP with very good performance is superior to the baseline methods.

A novel multi-model neuro-fuzzy-based MPPT for three-phase grid-connected photovoltaic system

DOE Office of Scientific and Technical Information (OSTI.GOV)

Chaouachi, Aymen; Kamel, Rashad M.; Nagasaka, Ken

This paper presents a novel methodology for Maximum Power Point Tracking (MPPT) of a grid-connected 20 kW photovoltaic (PV) system using neuro-fuzzy network. The proposed method predicts the reference PV voltage guarantying optimal power transfer between the PV generator and the main utility grid. The neuro-fuzzy network is composed of a fuzzy rule-based classifier and three multi-layered feed forwarded Artificial Neural Networks (ANN). Inputs of the network (irradiance and temperature) are classified before they are fed into the appropriated ANN for either training or estimation process while the output is the reference voltage. The main advantage of the proposed methodology,more » comparing to a conventional single neural network-based approach, is the distinct generalization ability regarding to the nonlinear and dynamic behavior of a PV generator. In fact, the neuro-fuzzy network is a neural network based multi-model machine learning that defines a set of local models emulating the complex and nonlinear behavior of a PV generator under a wide range of operating conditions. Simulation results under several rapid irradiance variations proved that the proposed MPPT method fulfilled the highest efficiency comparing to a conventional single neural network and the Perturb and Observe (P and O) algorithm dispositive. (author)« less

Hebbian based learning with winner-take-all for spiking neural networks

NASA Astrophysics Data System (ADS)

Gupta, Ankur; Long, Lyle

2009-03-01

Learning methods for spiking neural networks are not as well developed as the traditional neural networks that widely use back-propagation training. We propose and implement a Hebbian based learning method with winner-take-all competition for spiking neural networks. This approach is spike time dependent which makes it naturally well suited for a network of spiking neurons. Homeostasis with Hebbian learning is implemented which ensures stability and quicker learning. Homeostasis implies that the net sum of incoming weights associated with a neuron remains the same. Winner-take-all is also implemented for competitive learning between output neurons. We implemented this learning rule on a biologically based vision processing system that we are developing, and use layers of leaky integrate and fire neurons. The network when presented with 4 bars (or Gabor filters) of different orientation learns to recognize the bar orientations (or Gabor filters). After training, each output neuron learns to recognize a bar at specific orientation and responds by firing more vigorously to that bar and less vigorously to others. These neurons are found to have bell shaped tuning curves and are similar to the simple cells experimentally observed by Hubel and Wiesel in the striate cortex of cat and monkey.

NETS - A NEURAL NETWORK DEVELOPMENT TOOL, VERSION 3.0 (MACINTOSH VERSION)

NASA Technical Reports Server (NTRS)

Phillips, T. A.

1994-01-01

NETS, A Tool for the Development and Evaluation of Neural Networks, provides a simulation of Neural Network algorithms plus an environment for developing such algorithms. Neural Networks are a class of systems modeled after the human brain. Artificial Neural Networks are formed from hundreds or thousands of simulated neurons, connected to each other in a manner similar to brain neurons. Problems which involve pattern matching readily fit the class of problems which NETS is designed to solve. NETS uses the back propagation learning method for all of the networks which it creates. The nodes of a network are usually grouped together into clumps called layers. Generally, a network will have an input layer through which the various environment stimuli are presented to the network, and an output layer for determining the network's response. The number of nodes in these two layers is usually tied to some features of the problem being solved. Other layers, which form intermediate stops between the input and output layers, are called hidden layers. NETS allows the user to customize the patterns of connections between layers of a network. NETS also provides features for saving the weight values of a network during the learning process, which allows for more precise control over the learning process. NETS is an interpreter. Its method of execution is the familiar "read-evaluate-print" loop found in interpreted languages such as BASIC and LISP. The user is presented with a prompt which is the simulator's way of asking for input. After a command is issued, NETS will attempt to evaluate the command, which may produce more prompts requesting specific information or an error if the command is not understood. The typical process involved when using NETS consists of translating the problem into a format which uses input/output pairs, designing a network configuration for the problem, and finally training the network with input/output pairs until an acceptable error is reached. NETS allows the user to generate C code to implement the network loaded into the system. This permits the placement of networks as components, or subroutines, in other systems. In short, once a network performs satisfactorily, the Generate C Code option provides the means for creating a program separate from NETS to run the network. Other features: files may be stored in binary or ASCII format; multiple input propagation is permitted; bias values may be included; capability to scale data without writing scaling code; quick interactive testing of network from the main menu; and several options that allow the user to manipulate learning efficiency. NETS is written in ANSI standard C language to be machine independent. The Macintosh version (MSC-22108) includes code for both a graphical user interface version and a command line interface version. The machine independent version (MSC-21588) only includes code for the command line interface version of NETS 3.0. The Macintosh version requires a Macintosh II series computer and has been successfully implemented under System 7. Four executables are included on these diskettes, two for floating point operations and two for integer arithmetic. It requires Think C 5.0 to compile. A minimum of 1Mb of RAM is required for execution. Sample input files and executables for both the command line version and the Macintosh user interface version are provided on the distribution medium. The Macintosh version is available on a set of three 3.5 inch 800K Macintosh format diskettes. The machine independent version has been successfully implemented on an IBM PC series compatible running MS-DOS, a DEC VAX running VMS, a SunIPC running SunOS, and a CRAY Y-MP running UNICOS. Two executables for the IBM PC version are included on the MS-DOS distribution media, one compiled for floating point operations and one for integer arithmetic. The machine independent version is available on a set of three 5.25 inch 360K MS-DOS format diskettes (standard distribution medium) or a .25 inch streaming magnetic tape cartridge in UNIX tar format. NETS was developed in 1989 and updated in 1992. IBM PC is a registered trademark of International Business Machines. MS-DOS is a registered trademark of Microsoft Corporation. DEC, VAX, and VMS are trademarks of Digital Equipment Corporation. SunIPC and SunOS are trademarks of Sun Microsystems, Inc. CRAY Y-MP and UNICOS are trademarks of Cray Research, Inc.

NETS - A NEURAL NETWORK DEVELOPMENT TOOL, VERSION 3.0 (MACHINE INDEPENDENT VERSION)

NASA Technical Reports Server (NTRS)

Baffes, P. T.

1994-01-01

NETS, A Tool for the Development and Evaluation of Neural Networks, provides a simulation of Neural Network algorithms plus an environment for developing such algorithms. Neural Networks are a class of systems modeled after the human brain. Artificial Neural Networks are formed from hundreds or thousands of simulated neurons, connected to each other in a manner similar to brain neurons. Problems which involve pattern matching readily fit the class of problems which NETS is designed to solve. NETS uses the back propagation learning method for all of the networks which it creates. The nodes of a network are usually grouped together into clumps called layers. Generally, a network will have an input layer through which the various environment stimuli are presented to the network, and an output layer for determining the network's response. The number of nodes in these two layers is usually tied to some features of the problem being solved. Other layers, which form intermediate stops between the input and output layers, are called hidden layers. NETS allows the user to customize the patterns of connections between layers of a network. NETS also provides features for saving the weight values of a network during the learning process, which allows for more precise control over the learning process. NETS is an interpreter. Its method of execution is the familiar "read-evaluate-print" loop found in interpreted languages such as BASIC and LISP. The user is presented with a prompt which is the simulator's way of asking for input. After a command is issued, NETS will attempt to evaluate the command, which may produce more prompts requesting specific information or an error if the command is not understood. The typical process involved when using NETS consists of translating the problem into a format which uses input/output pairs, designing a network configuration for the problem, and finally training the network with input/output pairs until an acceptable error is reached. NETS allows the user to generate C code to implement the network loaded into the system. This permits the placement of networks as components, or subroutines, in other systems. In short, once a network performs satisfactorily, the Generate C Code option provides the means for creating a program separate from NETS to run the network. Other features: files may be stored in binary or ASCII format; multiple input propagation is permitted; bias values may be included; capability to scale data without writing scaling code; quick interactive testing of network from the main menu; and several options that allow the user to manipulate learning efficiency. NETS is written in ANSI standard C language to be machine independent. The Macintosh version (MSC-22108) includes code for both a graphical user interface version and a command line interface version. The machine independent version (MSC-21588) only includes code for the command line interface version of NETS 3.0. The Macintosh version requires a Macintosh II series computer and has been successfully implemented under System 7. Four executables are included on these diskettes, two for floating point operations and two for integer arithmetic. It requires Think C 5.0 to compile. A minimum of 1Mb of RAM is required for execution. Sample input files and executables for both the command line version and the Macintosh user interface version are provided on the distribution medium. The Macintosh version is available on a set of three 3.5 inch 800K Macintosh format diskettes. The machine independent version has been successfully implemented on an IBM PC series compatible running MS-DOS, a DEC VAX running VMS, a SunIPC running SunOS, and a CRAY Y-MP running UNICOS. Two executables for the IBM PC version are included on the MS-DOS distribution media, one compiled for floating point operations and one for integer arithmetic. The machine independent version is available on a set of three 5.25 inch 360K MS-DOS format diskettes (standard distribution medium) or a .25 inch streaming magnetic tape cartridge in UNIX tar format. NETS was developed in 1989 and updated in 1992. IBM PC is a registered trademark of International Business Machines. MS-DOS is a registered trademark of Microsoft Corporation. DEC, VAX, and VMS are trademarks of Digital Equipment Corporation. SunIPC and SunOS are trademarks of Sun Microsystems, Inc. CRAY Y-MP and UNICOS are trademarks of Cray Research, Inc.

Image quality assessment using deep convolutional networks

NASA Astrophysics Data System (ADS)

Li, Yezhou; Ye, Xiang; Li, Yong

2017-12-01

This paper proposes a method of accurately assessing image quality without a reference image by using a deep convolutional neural network. Existing training based methods usually utilize a compact set of linear filters for learning features of images captured by different sensors to assess their quality. These methods may not be able to learn the semantic features that are intimately related with the features used in human subject assessment. Observing this drawback, this work proposes training a deep convolutional neural network (CNN) with labelled images for image quality assessment. The ReLU in the CNN allows non-linear transformations for extracting high-level image features, providing a more reliable assessment of image quality than linear filters. To enable the neural network to take images of any arbitrary size as input, the spatial pyramid pooling (SPP) is introduced connecting the top convolutional layer and the fully-connected layer. In addition, the SPP makes the CNN robust to object deformations to a certain extent. The proposed method taking an image as input carries out an end-to-end learning process, and outputs the quality of the image. It is tested on public datasets. Experimental results show that it outperforms existing methods by a large margin and can accurately assess the image quality on images taken by different sensors of varying sizes.

Hourly air pollution concentrations and their important predictors over Houston, Texas using deep neural networks: case study of DISCOVER-AQ time period

NASA Astrophysics Data System (ADS)

Eslami, E.; Choi, Y.; Roy, A.

2017-12-01

Air quality forecasting carried out by chemical transport models often show significant error. This study uses a deep-learning approach over the Houston-Galveston-Brazoria (HGB) area to overcome this forecasting challenge, for the DISCOVER-AQ period (September 2013). Two approaches, deep neural network (DNN) using a Multi-Layer Perceptron (MLP) and Restricted Boltzmann Machine (RBM) were utilized. The proposed approaches analyzed input data by identifying features abstracted from its previous layer using a stepwise method. The approaches predicted hourly ozone and PM in September 2013 using several predictors of prior three days, including wind fields, temperature, relative humidity, cloud fraction, precipitation along with PM, ozone, and NOx concentrations. Model-measurement comparisons for available monitoring sites reported Indexes of Agreement (IOA) of around 0.95 for both DNN and RBM. A standard artificial neural network (ANN) (IOA=0.90) with similar architecture showed poorer performance than the deep networks, clearly demonstrating the superiority of the deep approaches. Additionally, each network (both deep and standard) performed significantly better than a previous CMAQ study, which showed an IOA of less than 0.80. The most influential input variables were identified using their associated weights, which represented the sensitivity of ozone to input parameters. The results indicate deep learning approaches can achieve more accurate ozone forecasting and identify the important input variables for ozone predictions in metropolitan areas.

Minimal perceptrons for memorizing complex patterns

NASA Astrophysics Data System (ADS)

Pastor, Marissa; Song, Juyong; Hoang, Danh-Tai; Jo, Junghyo

2016-11-01

Feedforward neural networks have been investigated to understand learning and memory, as well as applied to numerous practical problems in pattern classification. It is a rule of thumb that more complex tasks require larger networks. However, the design of optimal network architectures for specific tasks is still an unsolved fundamental problem. In this study, we consider three-layered neural networks for memorizing binary patterns. We developed a new complexity measure of binary patterns, and estimated the minimal network size for memorizing them as a function of their complexity. We formulated the minimal network size for regular, random, and complex patterns. In particular, the minimal size for complex patterns, which are neither ordered nor disordered, was predicted by measuring their Hamming distances from known ordered patterns. Our predictions agree with simulations based on the back-propagation algorithm.

A novel artificial neural network method for biomedical prediction based on matrix pseudo-inversion.

PubMed

Cai, Binghuang; Jiang, Xia

2014-04-01

Biomedical prediction based on clinical and genome-wide data has become increasingly important in disease diagnosis and classification. To solve the prediction problem in an effective manner for the improvement of clinical care, we develop a novel Artificial Neural Network (ANN) method based on Matrix Pseudo-Inversion (MPI) for use in biomedical applications. The MPI-ANN is constructed as a three-layer (i.e., input, hidden, and output layers) feed-forward neural network, and the weights connecting the hidden and output layers are directly determined based on MPI without a lengthy learning iteration. The LASSO (Least Absolute Shrinkage and Selection Operator) method is also presented for comparative purposes. Single Nucleotide Polymorphism (SNP) simulated data and real breast cancer data are employed to validate the performance of the MPI-ANN method via 5-fold cross validation. Experimental results demonstrate the efficacy of the developed MPI-ANN for disease classification and prediction, in view of the significantly superior accuracy (i.e., the rate of correct predictions), as compared with LASSO. The results based on the real breast cancer data also show that the MPI-ANN has better performance than other machine learning methods (including support vector machine (SVM), logistic regression (LR), and an iterative ANN). In addition, experiments demonstrate that our MPI-ANN could be used for bio-marker selection as well. Copyright © 2013 Elsevier Inc. All rights reserved.

STDP-based spiking deep convolutional neural networks for object recognition.

PubMed

Kheradpisheh, Saeed Reza; Ganjtabesh, Mohammad; Thorpe, Simon J; Masquelier, Timothée

2018-03-01

Previous studies have shown that spike-timing-dependent plasticity (STDP) can be used in spiking neural networks (SNN) to extract visual features of low or intermediate complexity in an unsupervised manner. These studies, however, used relatively shallow architectures, and only one layer was trainable. Another line of research has demonstrated - using rate-based neural networks trained with back-propagation - that having many layers increases the recognition robustness, an approach known as deep learning. We thus designed a deep SNN, comprising several convolutional (trainable with STDP) and pooling layers. We used a temporal coding scheme where the most strongly activated neurons fire first, and less activated neurons fire later or not at all. The network was exposed to natural images. Thanks to STDP, neurons progressively learned features corresponding to prototypical patterns that were both salient and frequent. Only a few tens of examples per category were required and no label was needed. After learning, the complexity of the extracted features increased along the hierarchy, from edge detectors in the first layer to object prototypes in the last layer. Coding was very sparse, with only a few thousands spikes per image, and in some cases the object category could be reasonably well inferred from the activity of a single higher-order neuron. More generally, the activity of a few hundreds of such neurons contained robust category information, as demonstrated using a classifier on Caltech 101, ETH-80, and MNIST databases. We also demonstrate the superiority of STDP over other unsupervised techniques such as random crops (HMAX) or auto-encoders. Taken together, our results suggest that the combination of STDP with latency coding may be a key to understanding the way that the primate visual system learns, its remarkable processing speed and its low energy consumption. These mechanisms are also interesting for artificial vision systems, particularly for hardware solutions. Copyright © 2017 Elsevier Ltd. All rights reserved.

Convective heat transfer and pressure drop of aqua based TiO2 nanofluids at different diameters of nanoparticles: Data analysis and modeling with artificial neural network

NASA Astrophysics Data System (ADS)

Hemmat Esfe, Mohammad; Nadooshan, Afshin Ahmadi; Arshi, Ali; Alirezaie, Ali

2018-03-01

In this study, experimental data related to the Nusselt number and pressure drop of aqueous nanofluids of Titania is modeled and estimated by using ANN with 2 hidden layers and 8 neurons in each layer. Also in this study the effect of various effective variables in the Nusselt number and pressure drop is surveyed. This study indicated that the neural network modeling has been able to model experimental data with great accuracy. The modeling regression coefficient for the data of Nusselt number and relative pressure drop is 99.94% and 99.97% respectively. Besides, it represented that the increment of the Reynolds number and concentration made the increment of Nusselt number and pressure drop of aqueous nanofluid.

Reservoir computing on the hypersphere

NASA Astrophysics Data System (ADS)

Andrecut, M.

Reservoir Computing (RC) refers to a Recurrent Neural Network (RNNs) framework, frequently used for sequence learning and time series prediction. The RC system consists of a random fixed-weight RNN (the input-hidden reservoir layer) and a classifier (the hidden-output readout layer). Here, we focus on the sequence learning problem, and we explore a different approach to RC. More specifically, we remove the nonlinear neural activation function, and we consider an orthogonal reservoir acting on normalized states on the unit hypersphere. Surprisingly, our numerical results show that the system’s memory capacity exceeds the dimensionality of the reservoir, which is the upper bound for the typical RC approach based on Echo State Networks (ESNs). We also show how the proposed system can be applied to symmetric cryptography problems, and we include a numerical implementation.

Volumetric multimodality neural network for brain tumor segmentation

NASA Astrophysics Data System (ADS)

Silvana Castillo, Laura; Alexandra Daza, Laura; Carlos Rivera, Luis; Arbeláez, Pablo

2017-11-01

Brain lesion segmentation is one of the hardest tasks to be solved in computer vision with an emphasis on the medical field. We present a convolutional neural network that produces a semantic segmentation of brain tumors, capable of processing volumetric data along with information from multiple MRI modalities at the same time. This results in the ability to learn from small training datasets and highly imbalanced data. Our method is based on DeepMedic, the state of the art in brain lesion segmentation. We develop a new architecture with more convolutional layers, organized in three parallel pathways with different input resolution, and additional fully connected layers. We tested our method over the 2015 BraTS Challenge dataset, reaching an average dice coefficient of 84%, while the standard DeepMedic implementation reached 74%.

Interactive learning in 2×2 normal form games by neural network agents

NASA Astrophysics Data System (ADS)

Spiliopoulos, Leonidas

2012-11-01

This paper models the learning process of populations of randomly rematched tabula rasa neural network (NN) agents playing randomly generated 2×2 normal form games of all strategic classes. This approach has greater external validity than the existing models in the literature, each of which is usually applicable to narrow subsets of classes of games (often a single game) and/or to fixed matching protocols. The learning prowess of NNs with hidden layers was impressive as they learned to play unique pure strategy equilibria with near certainty, adhered to principles of dominance and iterated dominance, and exhibited a preference for risk-dominant equilibria. In contrast, perceptron NNs were found to perform significantly worse than hidden layer NN agents and human subjects in experimental studies.

A feedforward artificial neural network based on quantum effect vector-matrix multipliers.

PubMed

Levy, H J; McGill, T C

1993-01-01

The vector-matrix multiplier is the engine of many artificial neural network implementations because it can simulate the way in which neurons collect weighted input signals from a dendritic arbor. A new technology for building analog weighting elements that is theoretically capable of densities and speeds far beyond anything that conventional VLSI in silicon could ever offer is presented. To illustrate the feasibility of such a technology, a small three-layer feedforward prototype network with five binary neurons and six tri-state synapses was built and used to perform all of the fundamental logic functions: XOR, AND, OR, and NOT.

Applying Gradient Descent in Convolutional Neural Networks

NASA Astrophysics Data System (ADS)

Cui, Nan

2018-04-01

With the development of the integrated circuit and computer science, people become caring more about solving practical issues via information technologies. Along with that, a new subject called Artificial Intelligent (AI) comes up. One popular research interest of AI is about recognition algorithm. In this paper, one of the most common algorithms, Convolutional Neural Networks (CNNs) will be introduced, for image recognition. Understanding its theory and structure is of great significance for every scholar who is interested in this field. Convolution Neural Network is an artificial neural network which combines the mathematical method of convolution and neural network. The hieratical structure of CNN provides it reliable computer speed and reasonable error rate. The most significant characteristics of CNNs are feature extraction, weight sharing and dimension reduction. Meanwhile, combining with the Back Propagation (BP) mechanism and the Gradient Descent (GD) method, CNNs has the ability to self-study and in-depth learning. Basically, BP provides an opportunity for backwardfeedback for enhancing reliability and GD is used for self-training process. This paper mainly discusses the CNN and the related BP and GD algorithms, including the basic structure and function of CNN, details of each layer, the principles and features of BP and GD, and some examples in practice with a summary in the end.

An Indoor Positioning Technique Based on a Feed-Forward Artificial Neural Network Using Levenberg-Marquardt Learning Method

NASA Astrophysics Data System (ADS)

Pahlavani, P.; Gholami, A.; Azimi, S.

2017-09-01

This paper presents an indoor positioning technique based on a multi-layer feed-forward (MLFF) artificial neural networks (ANN). Most of the indoor received signal strength (RSS)-based WLAN positioning systems use the fingerprinting technique that can be divided into two phases: the offline (calibration) phase and the online (estimation) phase. In this paper, RSSs were collected for all references points in four directions and two periods of time (Morning and Evening). Hence, RSS readings were sampled at a regular time interval and specific orientation at each reference point. The proposed ANN based model used Levenberg-Marquardt algorithm for learning and fitting the network to the training data. This RSS readings in all references points and the known position of these references points was prepared for training phase of the proposed MLFF neural network. Eventually, the average positioning error for this network using 30% check and validation data was computed approximately 2.20 meter.

Macrocell path loss prediction using artificial intelligence techniques

NASA Astrophysics Data System (ADS)

Usman, Abraham U.; Okereke, Okpo U.; Omizegba, Elijah E.

2014-04-01

The prediction of propagation loss is a practical non-linear function approximation problem which linear regression or auto-regression models are limited in their ability to handle. However, some computational Intelligence techniques such as artificial neural networks (ANNs) and adaptive neuro-fuzzy inference systems (ANFISs) have been shown to have great ability to handle non-linear function approximation and prediction problems. In this study, the multiple layer perceptron neural network (MLP-NN), radial basis function neural network (RBF-NN) and an ANFIS network were trained using actual signal strength measurement taken at certain suburban areas of Bauchi metropolis, Nigeria. The trained networks were then used to predict propagation losses at the stated areas under differing conditions. The predictions were compared with the prediction accuracy of the popular Hata model. It was observed that ANFIS model gave a better fit in all cases having higher R2 values in each case and on average is more robust than MLP and RBF models as it generalises better to a different data.

Multispectral embedding-based deep neural network for three-dimensional human pose recovery

NASA Astrophysics Data System (ADS)

Yu, Jialin; Sun, Jifeng

2018-01-01

Monocular image-based three-dimensional (3-D) human pose recovery aims to retrieve 3-D poses using the corresponding two-dimensional image features. Therefore, the pose recovery performance highly depends on the image representations. We propose a multispectral embedding-based deep neural network (MSEDNN) to automatically obtain the most discriminative features from multiple deep convolutional neural networks and then embed their penultimate fully connected layers into a low-dimensional manifold. This compact manifold can explore not only the optimum output from multiple deep networks but also the complementary properties of them. Furthermore, the distribution of each hierarchy discriminative manifold is sufficiently smooth so that the training process of our MSEDNN can be effectively implemented only using few labeled data. Our proposed network contains a body joint detector and a human pose regressor that are jointly trained. Extensive experiments conducted on four databases show that our proposed MSEDNN can achieve the best recovery performance compared with the state-of-the-art methods.

Predicting outcomes in patients with perforated gastroduodenal ulcers: artificial neural network modelling indicates a highly complex disease.

PubMed

Søreide, K; Thorsen, K; Søreide, J A

2015-02-01

Mortality prediction models for patients with perforated peptic ulcer (PPU) have not yielded consistent or highly accurate results. Given the complex nature of this disease, which has many non-linear associations with outcomes, we explored artificial neural networks (ANNs) to predict the complex interactions between the risk factors of PPU and death among patients with this condition. ANN modelling using a standard feed-forward, back-propagation neural network with three layers (i.e., an input layer, a hidden layer and an output layer) was used to predict the 30-day mortality of consecutive patients from a population-based cohort undergoing surgery for PPU. A receiver-operating characteristic (ROC) analysis was used to assess model accuracy. Of the 172 patients, 168 had their data included in the model; the data of 117 (70%) were used for the training set, and the data of 51 (39%) were used for the test set. The accuracy, as evaluated by area under the ROC curve (AUC), was best for an inclusive, multifactorial ANN model (AUC 0.90, 95% CIs 0.85-0.95; p < 0.001). This model outperformed standard predictive scores, including Boey and PULP. The importance of each variable decreased as the number of factors included in the ANN model increased. The prediction of death was most accurate when using an ANN model with several univariate influences on the outcome. This finding demonstrates that PPU is a highly complex disease for which clinical prognoses are likely difficult. The incorporation of computerised learning systems might enhance clinical judgments to improve decision making and outcome prediction.

Squeeze-SegNet: a new fast deep convolutional neural network for semantic segmentation

NASA Astrophysics Data System (ADS)

Nanfack, Geraldin; Elhassouny, Azeddine; Oulad Haj Thami, Rachid

2018-04-01

The recent researches in Deep Convolutional Neural Network have focused their attention on improving accuracy that provide significant advances. However, if they were limited to classification tasks, nowadays with contributions from Scientific Communities who are embarking in this field, they have become very useful in higher level tasks such as object detection and pixel-wise semantic segmentation. Thus, brilliant ideas in the field of semantic segmentation with deep learning have completed the state of the art of accuracy, however this architectures become very difficult to apply in embedded systems as is the case for autonomous driving. We present a new Deep fully Convolutional Neural Network for pixel-wise semantic segmentation which we call Squeeze-SegNet. The architecture is based on Encoder-Decoder style. We use a SqueezeNet-like encoder and a decoder formed by our proposed squeeze-decoder module and upsample layer using downsample indices like in SegNet and we add a deconvolution layer to provide final multi-channel feature map. On datasets like Camvid or City-states, our net gets SegNet-level accuracy with less than 10 times fewer parameters than SegNet.

A Novel Extreme Learning Machine Classification Model for e-Nose Application Based on the Multiple Kernel Approach.

PubMed

Jian, Yulin; Huang, Daoyu; Yan, Jia; Lu, Kun; Huang, Ying; Wen, Tailai; Zeng, Tanyue; Zhong, Shijie; Xie, Qilong

2017-06-19

A novel classification model, named the quantum-behaved particle swarm optimization (QPSO)-based weighted multiple kernel extreme learning machine (QWMK-ELM), is proposed in this paper. Experimental validation is carried out with two different electronic nose (e-nose) datasets. Being different from the existing multiple kernel extreme learning machine (MK-ELM) algorithms, the combination coefficients of base kernels are regarded as external parameters of single-hidden layer feedforward neural networks (SLFNs). The combination coefficients of base kernels, the model parameters of each base kernel, and the regularization parameter are optimized by QPSO simultaneously before implementing the kernel extreme learning machine (KELM) with the composite kernel function. Four types of common single kernel functions (Gaussian kernel, polynomial kernel, sigmoid kernel, and wavelet kernel) are utilized to constitute different composite kernel functions. Moreover, the method is also compared with other existing classification methods: extreme learning machine (ELM), kernel extreme learning machine (KELM), k-nearest neighbors (KNN), support vector machine (SVM), multi-layer perceptron (MLP), radical basis function neural network (RBFNN), and probabilistic neural network (PNN). The results have demonstrated that the proposed QWMK-ELM outperforms the aforementioned methods, not only in precision, but also in efficiency for gas classification.

Recruitment and Consolidation of Cell Assemblies for Words by Way of Hebbian Learning and Competition in a Multi-Layer Neural Network

PubMed Central

Garagnani, Max; Wennekers, Thomas; Pulvermüller, Friedemann

2009-01-01

Current cognitive theories postulate either localist representations of knowledge or fully overlapping, distributed ones. We use a connectionist model that closely replicates known anatomical properties of the cerebral cortex and neurophysiological principles to show that Hebbian learning in a multi-layer neural network leads to memory traces (cell assemblies) that are both distributed and anatomically distinct. Taking the example of word learning based on action-perception correlation, we document mechanisms underlying the emergence of these assemblies, especially (i) the recruitment of neurons and consolidation of connections defining the kernel of the assembly along with (ii) the pruning of the cell assembly’s halo (consisting of very weakly connected cells). We found that, whereas a learning rule mapping covariance led to significant overlap and merging of assemblies, a neurobiologically grounded synaptic plasticity rule with fixed LTP/LTD thresholds produced minimal overlap and prevented merging, exhibiting competitive learning behaviour. Our results are discussed in light of current theories of language and memory. As simulations with neurobiologically realistic neural networks demonstrate here spontaneous emergence of lexical representations that are both cortically dispersed and anatomically distinct, both localist and distributed cognitive accounts receive partial support. PMID:20396612

Recruitment and Consolidation of Cell Assemblies for Words by Way of Hebbian Learning and Competition in a Multi-Layer Neural Network.

PubMed

Garagnani, Max; Wennekers, Thomas; Pulvermüller, Friedemann

2009-06-01

Current cognitive theories postulate either localist representations of knowledge or fully overlapping, distributed ones. We use a connectionist model that closely replicates known anatomical properties of the cerebral cortex and neurophysiological principles to show that Hebbian learning in a multi-layer neural network leads to memory traces (cell assemblies) that are both distributed and anatomically distinct. Taking the example of word learning based on action-perception correlation, we document mechanisms underlying the emergence of these assemblies, especially (i) the recruitment of neurons and consolidation of connections defining the kernel of the assembly along with (ii) the pruning of the cell assembly's halo (consisting of very weakly connected cells). We found that, whereas a learning rule mapping covariance led to significant overlap and merging of assemblies, a neurobiologically grounded synaptic plasticity rule with fixed LTP/LTD thresholds produced minimal overlap and prevented merging, exhibiting competitive learning behaviour. Our results are discussed in light of current theories of language and memory. As simulations with neurobiologically realistic neural networks demonstrate here spontaneous emergence of lexical representations that are both cortically dispersed and anatomically distinct, both localist and distributed cognitive accounts receive partial support.

Optimal and robust control of a class of nonlinear systems using dynamically re-optimised single network adaptive critic design

NASA Astrophysics Data System (ADS)

Tiwari, Shivendra N.; Padhi, Radhakant

2018-01-01

Following the philosophy of adaptive optimal control, a neural network-based state feedback optimal control synthesis approach is presented in this paper. First, accounting for a nominal system model, a single network adaptive critic (SNAC) based multi-layered neural network (called as NN1) is synthesised offline. However, another linear-in-weight neural network (called as NN2) is trained online and augmented to NN1 in such a manner that their combined output represent the desired optimal costate for the actual plant. To do this, the nominal model needs to be updated online to adapt to the actual plant, which is done by synthesising yet another linear-in-weight neural network (called as NN3) online. Training of NN3 is done by utilising the error information between the nominal and actual states and carrying out the necessary Lyapunov stability analysis using a Sobolev norm based Lyapunov function. This helps in training NN2 successfully to capture the required optimal relationship. The overall architecture is named as 'Dynamically Re-optimised single network adaptive critic (DR-SNAC)'. Numerical results for two motivating illustrative problems are presented, including comparison studies with closed form solution for one problem, which clearly demonstrate the effectiveness and benefit of the proposed approach.

Neural networks with local receptive fields and superlinear VC dimension.

PubMed

Schmitt, Michael

2002-04-01

Local receptive field neurons comprise such well-known and widely used unit types as radial basis function (RBF) neurons and neurons with center-surround receptive field. We study the Vapnik-Chervonenkis (VC) dimension of feedforward neural networks with one hidden layer of these units. For several variants of local receptive field neurons, we show that the VC dimension of these networks is superlinear. In particular, we establish the bound Omega(W log k) for any reasonably sized network with W parameters and k hidden nodes. This bound is shown to hold for discrete center-surround receptive field neurons, which are physiologically relevant models of cells in the mammalian visual system, for neurons computing a difference of gaussians, which are popular in computational vision, and for standard RBF neurons, a major alternative to sigmoidal neurons in artificial neural networks. The result for RBF neural networks is of particular interest since it answers a question that has been open for several years. The results also give rise to lower bounds for networks with fixed input dimension. Regarding constants, all bounds are larger than those known thus far for similar architectures with sigmoidal neurons. The superlinear lower bounds contrast with linear upper bounds for single local receptive field neurons also derived here.

Proposal for an All-Spin Artificial Neural Network: Emulating Neural and Synaptic Functionalities Through Domain Wall Motion in Ferromagnets.

PubMed

Sengupta, Abhronil; Shim, Yong; Roy, Kaushik

2016-12-01

Non-Boolean computing based on emerging post-CMOS technologies can potentially pave the way for low-power neural computing platforms. However, existing work on such emerging neuromorphic architectures have either focused on solely mimicking the neuron, or the synapse functionality. While memristive devices have been proposed to emulate biological synapses, spintronic devices have proved to be efficient at performing the thresholding operation of the neuron at ultra-low currents. In this work, we propose an All-Spin Artificial Neural Network where a single spintronic device acts as the basic building block of the system. The device offers a direct mapping to synapse and neuron functionalities in the brain while inter-layer network communication is accomplished via CMOS transistors. To the best of our knowledge, this is the first demonstration of a neural architecture where a single nanoelectronic device is able to mimic both neurons and synapses. The ultra-low voltage operation of low resistance magneto-metallic neurons enables the low-voltage operation of the array of spintronic synapses, thereby leading to ultra-low power neural architectures. Device-level simulations, calibrated to experimental results, was used to drive the circuit and system level simulations of the neural network for a standard pattern recognition problem. Simulation studies indicate energy savings by  ∼  100× in comparison to a corresponding digital/analog CMOS neuron implementation.

Emergent latent symbol systems in recurrent neural networks

NASA Astrophysics Data System (ADS)

Monner, Derek; Reggia, James A.

2012-12-01

Fodor and Pylyshyn [(1988). Connectionism and cognitive architecture: A critical analysis. Cognition, 28(1-2), 3-71] famously argued that neural networks cannot behave systematically short of implementing a combinatorial symbol system. A recent response from Frank et al. [(2009). Connectionist semantic systematicity. Cognition, 110(3), 358-379] claimed to have trained a neural network to behave systematically without implementing a symbol system and without any in-built predisposition towards combinatorial representations. We believe systems like theirs may in fact implement a symbol system on a deeper and more interesting level: one where the symbols are latent - not visible at the level of network structure. In order to illustrate this possibility, we demonstrate our own recurrent neural network that learns to understand sentence-level language in terms of a scene. We demonstrate our model's learned understanding by testing it on novel sentences and scenes. By paring down our model into an architecturally minimal version, we demonstrate how it supports combinatorial computation over distributed representations by using the associative memory operations of Vector Symbolic Architectures. Knowledge of the model's memory scheme gives us tools to explain its errors and construct superior future models. We show how the model designs and manipulates a latent symbol system in which the combinatorial symbols are patterns of activation distributed across the layers of a neural network, instantiating a hybrid of classical symbolic and connectionist representations that combines advantages of both.

Performance Evaluation of Public Non-Profit Hospitals Using a BP Artificial Neural Network: The Case of Hubei Province in China

PubMed Central

Li, Chunhui; Yu, Chuanhua

2013-01-01

To provide a reference for evaluating public non-profit hospitals in the new environment of medical reform, we established a performance evaluation system for public non-profit hospitals. The new “input-output” performance model for public non-profit hospitals is based on four primary indexes (input, process, output and effect) that include 11 sub-indexes and 41 items. The indicator weights were determined using the analytic hierarchy process (AHP) and entropy weight method. The BP neural network was applied to evaluate the performance of 14 level-3 public non-profit hospitals located in Hubei Province. The most stable BP neural network was produced by comparing different numbers of neurons in the hidden layer and using the “Leave-one-out” Cross Validation method. The performance evaluation system we established for public non-profit hospitals could reflect the basic goal of the new medical health system reform in China. Compared with PLSR, the result indicated that the BP neural network could be used effectively for evaluating the performance public non-profit hospitals. PMID:23955238

Automated Analysis of Planktic Foraminifers Part III: Neural Network Classification

NASA Astrophysics Data System (ADS)

Schiebel, R.; Bollmann, J.; Quinn, P.; Vela, M.; Schmidt, D. N.; Thierstein, H. R.

2003-04-01

The abundance and assemblage composition of microplankton, together with the chemical and stable isotopic composition of their shells, are among the most successful methods in paleoceanography and paleoclimatology. However, the manual collection of statistically significant numbers of unbiased, reproducible data is time consuming. Consequently, automated microfossil analysis and species recognition has been a long-standing goal in micropaleontology. We have developed a Windows based software package COGNIS for the segmentation, preprocessing, and classification of automatically acquired microfossil images (see Part II, Bollmann et al., this volume), using operator designed neural network structures. With a five-layered convolutional neural network we obtain an average recognition rate of 75 % (max. 88 %) for 6 taxa (N. dutertrei, N. pachyderma dextral, N. pachyderma sinistral, G. inflata, G. menardii/tumida, O. universa), represented by 50 images each for 20 classes (separation of spiral and umbilical views, and of sinistral and dextral forms). Our investigation indicates that neural networks hold great potential for the automated classification of planktic foraminifers and offer new perspectives in micropaleontology, paleoceanography, and paleoclimatology (see Part I, Schmidt et al., this volume).

Real-time determination of fringe pattern frequencies: An application to pressure measurement

NASA Astrophysics Data System (ADS)

Sciammarella, Cesar A.; Piroozan, Parham

2007-05-01

Retrieving information in real time from fringe patterns is a topic of a great deal of interest in scientific and engineering applications of optical methods. This paper presents a method for fringe frequency determination based on the capability of neural networks to recognize signals that are similar but not identical to signals used to train the neural network. Sampled patterns are generated by calibration and stored in memory. Incoming patterns are analyzed by a back-propagation neural network at the speed of the recording device, a CCD camera. This method of information retrieval is utilized to measure pressures on a boundary layer flow. The sensor combines optics and electronics to analyze dynamic pressure distributions and to feed information to a control system that is capable to preserve the stability of the flow.

Reformulated Neural Network (ReNN): a New Alternative for Data-driven Modelling in Hydrology and Water Resources Engineering

NASA Astrophysics Data System (ADS)

Razavi, S.; Tolson, B.; Burn, D.; Seglenieks, F.

2012-04-01

Reformulated Neural Network (ReNN) has been recently developed as an efficient and more effective alternative to feedforward multi-layer perceptron (MLP) neural networks [Razavi, S., and Tolson, B. A. (2011). "A new formulation for feedforward neural networks." IEEE Transactions on Neural Networks, 22(10), 1588-1598, DOI: 1510.1109/TNN.2011.2163169]. This presentation initially aims to introduce the ReNN to the water resources community and then demonstrates ReNN applications to water resources related problems. ReNN is essentially equivalent to a single-hidden-layer MLP neural network but defined on a new set of network variables which is more effective than the traditional set of network weights and biases. The main features of the new network variables are that they are geometrically interpretable and each variable has a distinct role in forming the network response. ReNN is more efficiently trained as it has a less complex error response surface. In addition to the ReNN training efficiency, the interpretability of the ReNN variables enables the users to monitor and understand the internal behaviour of the network while training. Regularization in the ReNN response can be also directly measured and controlled. This feature improves the generalization ability of the network. The appeal of the ReNN is demonstrated with two ReNN applications to water resources engineering problems. In the first application, the ReNN is used to model the rainfall-runoff relationships in multiple watersheds in the Great Lakes basin located in northeastern North America. Modelling inflows to the Great Lakes are of great importance to the management of the Great Lakes system. Due to the lack of some detailed physical data about existing control structures in many subwatersheds of this huge basin, the data-driven approach to modelling such as the ReNN are required to replace predictions from a physically-based rainfall runoff model. Unlike traditional MLPs, the ReNN does not necessarily require an independent set of data for validation as the ReNN has the capability to control and verify the network degree of regularization. As such, the ReNN can be very beneficial in this case study as the data available in this case study is limited. In the second application, ReNN is fitted on the response function of the SWAT hydrologic model to act as a fast-to-run response surface surrogate (i.e., metamodel) of the original computationally intensive SWAT model. Besides the training efficiency gains, the ReNN applications demonstrate how the ReNN interpretability could help users develop more reliable networks which perform predictably better in terms of generalization.

Decoding the cortical transformations for visually guided reaching in 3D space.

PubMed

Blohm, Gunnar; Keith, Gerald P; Crawford, J Douglas

2009-06-01

To explore the possible cortical mechanisms underlying the 3-dimensional (3D) visuomotor transformation for reaching, we trained a 4-layer feed-forward artificial neural network to compute a reach vector (output) from the visual positions of both the hand and target viewed from different eye and head orientations (inputs). The emergent properties of the intermediate layers reflected several known neurophysiological findings, for example, gain field-like modulations and position-dependent shifting of receptive fields (RFs). We performed a reference frame analysis for each individual network unit, simulating standard electrophysiological experiments, that is, RF mapping (unit input), motor field mapping, and microstimulation effects (unit outputs). At the level of individual units (in both intermediate layers), the 3 different electrophysiological approaches identified different reference frames, demonstrating that these techniques reveal different neuronal properties and suggesting that a comparison across these techniques is required to understand the neural code of physiological networks. This analysis showed fixed input-output relationships within each layer and, more importantly, within each unit. These local reference frame transformation modules provide the basic elements for the global transformation; their parallel contributions are combined in a gain field-like fashion at the population level to implement both the linear and nonlinear elements of the 3D visuomotor transformation.

Bias correction of temperature produced by the Community Climate System Model using Artificial Neural Networks

NASA Astrophysics Data System (ADS)

Moghim, S.; Hsu, K.; Bras, R. L.

2013-12-01

General Circulation Models (GCMs) are used to predict circulation and energy transfers between the atmosphere and the land. It is known that these models produce biased results that will have impact on their uses. This work proposes a new method for bias correction: the equidistant cumulative distribution function-artificial neural network (EDCDFANN) procedure. The method uses artificial neural networks (ANNs) as a surrogate model to estimate bias-corrected temperature, given an identification of the system derived from GCM models output variables. A two-layer feed forward neural network is trained with observations during a historical period and then the adjusted network can be used to predict bias-corrected temperature for future periods. To capture the extreme values this method is combined with the equidistant CDF matching method (EDCDF, Li et al. 2010). The proposed method is tested with the Community Climate System Model (CCSM3) outputs using air and skin temperature, specific humidity, shortwave and longwave radiation as inputs to the ANN. This method decreases the mean square error and increases the spatial correlation between the modeled temperature and the observed one. The results indicate the EDCDFANN has potential to remove the biases of the model outputs.

Forecasting outpatient visits using empirical mode decomposition coupled with back-propagation artificial neural networks optimized by particle swarm optimization

PubMed Central

Huang, Daizheng; Wu, Zhihui

2017-01-01

Accurately predicting the trend of outpatient visits by mathematical modeling can help policy makers manage hospitals effectively, reasonably organize schedules for human resources and finances, and appropriately distribute hospital material resources. In this study, a hybrid method based on empirical mode decomposition and back-propagation artificial neural networks optimized by particle swarm optimization is developed to forecast outpatient visits on the basis of monthly numbers. The data outpatient visits are retrieved from January 2005 to December 2013 and first obtained as the original time series. Second, the original time series is decomposed into a finite and often small number of intrinsic mode functions by the empirical mode decomposition technique. Third, a three-layer back-propagation artificial neural network is constructed to forecast each intrinsic mode functions. To improve network performance and avoid falling into a local minimum, particle swarm optimization is employed to optimize the weights and thresholds of back-propagation artificial neural networks. Finally, the superposition of forecasting results of the intrinsic mode functions is regarded as the ultimate forecasting value. Simulation indicates that the proposed method attains a better performance index than the other four methods. PMID:28222194

Forecasting outpatient visits using empirical mode decomposition coupled with back-propagation artificial neural networks optimized by particle swarm optimization.

PubMed

Huang, Daizheng; Wu, Zhihui

2017-01-01

Accurately predicting the trend of outpatient visits by mathematical modeling can help policy makers manage hospitals effectively, reasonably organize schedules for human resources and finances, and appropriately distribute hospital material resources. In this study, a hybrid method based on empirical mode decomposition and back-propagation artificial neural networks optimized by particle swarm optimization is developed to forecast outpatient visits on the basis of monthly numbers. The data outpatient visits are retrieved from January 2005 to December 2013 and first obtained as the original time series. Second, the original time series is decomposed into a finite and often small number of intrinsic mode functions by the empirical mode decomposition technique. Third, a three-layer back-propagation artificial neural network is constructed to forecast each intrinsic mode functions. To improve network performance and avoid falling into a local minimum, particle swarm optimization is employed to optimize the weights and thresholds of back-propagation artificial neural networks. Finally, the superposition of forecasting results of the intrinsic mode functions is regarded as the ultimate forecasting value. Simulation indicates that the proposed method attains a better performance index than the other four methods.

What is the optimal architecture for visual information routing?

PubMed

Wolfrum, Philipp; von der Malsburg, Christoph

2007-12-01

Analyzing the design of networks for visual information routing is an underconstrained problem due to insufficient anatomical and physiological data. We propose here optimality criteria for the design of routing networks. For a very general architecture, we derive the number of routing layers and the fanout that minimize the required neural circuitry. The optimal fanout l is independent of network size, while the number k of layers scales logarithmically (with a prefactor below 1), with the number n of visual resolution units to be routed independently. The results are found to agree with data of the primate visual system.

A Hybrid FEM-ANN Approach for Slope Instability Prediction

NASA Astrophysics Data System (ADS)

Verma, A. K.; Singh, T. N.; Chauhan, Nikhil Kumar; Sarkar, K.

2016-09-01

Assessment of slope stability is one of the most critical aspects for the life of a slope. In any slope vulnerability appraisal, Factor Of Safety (FOS) is the widely accepted index to understand, how close or far a slope from the failure. In this work, an attempt has been made to simulate a road cut slope in a landslide prone area in Rudrapryag, Uttarakhand, India which lies near Himalayan geodynamic mountain belt. A combination of Finite Element Method (FEM) and Artificial Neural Network (ANN) has been adopted to predict FOS of the slope. In ANN, a three layer, feed- forward back-propagation neural network with one input layer and one hidden layer with three neurons and one output layer has been considered and trained using datasets generated from numerical analysis of the slope and validated with new set of field slope data. Mean absolute percentage error estimated as 1.04 with coefficient of correlation between the FOS of FEM and ANN as 0.973, which indicates that the system is very vigorous and fast to predict FOS for any slope.

Classification of amyotrophic lateral sclerosis disease based on convolutional neural network and reinforcement sample learning algorithm.

PubMed

Sengur, Abdulkadir; Akbulut, Yaman; Guo, Yanhui; Bajaj, Varun

2017-12-01

Electromyogram (EMG) signals contain useful information of the neuromuscular diseases like amyotrophic lateral sclerosis (ALS). ALS is a well-known brain disease, which can progressively degenerate the motor neurons. In this paper, we propose a deep learning based method for efficient classification of ALS and normal EMG signals. Spectrogram, continuous wavelet transform (CWT), and smoothed pseudo Wigner-Ville distribution (SPWVD) have been employed for time-frequency (T-F) representation of EMG signals. A convolutional neural network is employed to classify these features. In it, Two convolution layers, two pooling layer, a fully connected layer and a lost function layer is considered in CNN architecture. The CNN architecture is trained with the reinforcement sample learning strategy. The efficiency of the proposed implementation is tested on publicly available EMG dataset. The dataset contains 89 ALS and 133 normal EMG signals with 24 kHz sampling frequency. Experimental results show 96.80% accuracy. The obtained results are also compared with other methods, which show the superiority of the proposed method.

Multi-focus image fusion with the all convolutional neural network

NASA Astrophysics Data System (ADS)

Du, Chao-ben; Gao, She-sheng

2018-01-01

A decision map contains complete and clear information about the image to be fused, which is crucial to various image fusion issues, especially multi-focus image fusion. However, in order to get a satisfactory image fusion effect, getting a decision map is very necessary and usually difficult to finish. In this letter, we address this problem with convolutional neural network (CNN), aiming to get a state-of-the-art decision map. The main idea is that the max-pooling of CNN is replaced by a convolution layer, the residuals are propagated backwards by gradient descent, and the training parameters of the individual layers of the CNN are updated layer by layer. Based on this, we propose a new all CNN (ACNN)-based multi-focus image fusion method in spatial domain. We demonstrate that the decision map obtained from the ACNN is reliable and can lead to high-quality fusion results. Experimental results clearly validate that the proposed algorithm can obtain state-of-the-art fusion performance in terms of both qualitative and quantitative evaluations.

Real-Time Adaptive Color Segmentation by Neural Networks

NASA Technical Reports Server (NTRS)

Duong, Tuan A.

2004-01-01

Artificial neural networks that would utilize the cascade error projection (CEP) algorithm have been proposed as means of autonomous, real-time, adaptive color segmentation of images that change with time. In the original intended application, such a neural network would be used to analyze digitized color video images of terrain on a remote planet as viewed from an uninhabited spacecraft approaching the planet. During descent toward the surface of the planet, information on the segmentation of the images into differently colored areas would be updated adaptively in real time to capture changes in contrast, brightness, and resolution, all in an effort to identify a safe and scientifically productive landing site and provide control feedback to steer the spacecraft toward that site. Potential terrestrial applications include monitoring images of crops to detect insect invasions and monitoring of buildings and other facilities to detect intruders. The CEP algorithm is reliable and is well suited to implementation in very-large-scale integrated (VLSI) circuitry. It was chosen over other neural-network learning algorithms because it is better suited to realtime learning: It provides a self-evolving neural-network structure, requires fewer iterations to converge and is more tolerant to low resolution (that is, fewer bits) in the quantization of neural-network synaptic weights. Consequently, a CEP neural network learns relatively quickly, and the circuitry needed to implement it is relatively simple. Like other neural networks, a CEP neural network includes an input layer, hidden units, and output units (see figure). As in other neural networks, a CEP network is presented with a succession of input training patterns, giving rise to a set of outputs that are compared with the desired outputs. Also as in other neural networks, the synaptic weights are updated iteratively in an effort to bring the outputs closer to target values. A distinctive feature of the CEP neural network and algorithm is that each update of synaptic weights takes place in conjunction with the addition of another hidden unit, which then remains in place as still other hidden units are added on subsequent iterations. For a given training pattern, the synaptic weight between (1) the inputs and the previously added hidden units and (2) the newly added hidden unit is updated by an amount proportional to the partial derivative of a quadratic error function with respect to the synaptic weight. The synaptic weight between the newly added hidden unit and each output unit is given by a more complex function that involves the errors between the outputs and their target values, the transfer functions (hyperbolic tangents) of the neural units, and the derivatives of the transfer functions.

Network architectures and circuit function: testing alternative hypotheses in multifunctional networks.

PubMed

Leonard, J L

2000-05-01

Understanding how species-typical movement patterns are organized in the nervous system is a central question in neurobiology. The current explanations involve 'alphabet' models in which an individual neuron may participate in the circuit for several behaviors but each behavior is specified by a specific neural circuit. However, not all of the well-studied model systems fit the 'alphabet' model. The 'equation' model provides an alternative possibility, whereby a system of parallel motor neurons, each with a unique (but overlapping) field of innervation, can account for the production of stereotyped behavior patterns by variable circuits. That is, it is possible for such patterns to arise as emergent properties of a generalized neural network in the absence of feedback, a simple version of a 'self-organizing' behavioral system. Comparison of systems of identified neurons suggest that the 'alphabet' model may account for most observations where CPGs act to organize motor patterns. Other well-known model systems, involving architectures corresponding to feed-forward neural networks with a hidden layer, may organize patterned behavior in a manner consistent with the 'equation' model. Such architectures are found in the Mauthner and reticulospinal circuits, 'escape' locomotion in cockroaches, CNS control of Aplysia gill, and may also be important in the coordination of sensory information and motor systems in insect mushroom bodies and the vertebrate hippocampus. The hidden layer of such networks may serve as an 'internal representation' of the behavioral state and/or body position of the animal, allowing the animal to fine-tune oriented, or particularly context-sensitive, movements to the prevalent conditions. Experiments designed to distinguish between the two models in cases where they make mutually exclusive predictions provide an opportunity to elucidate the neural mechanisms by which behavior is organized in vivo and in vitro. Copyright 2000 S. Karger AG, Basel

Hierarchical neural network model of the visual system determining figure/ground relation

NASA Astrophysics Data System (ADS)

Kikuchi, Masayuki

2017-07-01

One of the most important functions of the visual perception in the brain is figure/ground interpretation from input images. Figural region in 2D image corresponding to object in 3D space are distinguished from background region extended behind the object. Previously the author proposed a neural network model of figure/ground separation constructed on the standpoint that local geometric features such as curvatures and outer angles at corners are extracted and propagated along input contour in a single layer network (Kikuchi & Akashi, 2001). However, such a processing principle has the defect that signal propagation requires manyiterations despite the fact that actual visual system determines figure/ground relation within the short period (Zhou et al., 2000). In order to attain speed-up for determining figure/ground, this study incorporates hierarchical architecture into the previous model. This study confirmed the effect of the hierarchization as for the computation time by simulation. As the number of layers increased, the required computation time reduced. However, such speed-up effect was saturatedas the layers increased to some extent. This study attempted to explain this saturation effect by the notion of average distance between vertices in the area of complex network, and succeeded to mimic the saturation effect by computer simulation.

Spatial and Time Domain Feature of ERP Speller System Extracted via Convolutional Neural Network.

PubMed

Yoon, Jaehong; Lee, Jungnyun; Whang, Mincheol

2018-01-01

Feature of event-related potential (ERP) has not been completely understood and illiteracy problem remains unsolved. To this end, P300 peak has been used as the feature of ERP in most brain-computer interface applications, but subjects who do not show such peak are common. Recent development of convolutional neural network provides a way to analyze spatial and temporal features of ERP. Here, we train the convolutional neural network with 2 convolutional layers whose feature maps represented spatial and temporal features of event-related potential. We have found that nonilliterate subjects' ERP show high correlation between occipital lobe and parietal lobe, whereas illiterate subjects only show correlation between neural activities from frontal lobe and central lobe. The nonilliterates showed peaks in P300, P500, and P700, whereas illiterates mostly showed peaks in around P700. P700 was strong in both subjects. We found that P700 peak may be the key feature of ERP as it appears in both illiterate and nonilliterate subjects.

Spatial and Time Domain Feature of ERP Speller System Extracted via Convolutional Neural Network

PubMed Central

2018-01-01

Feature of event-related potential (ERP) has not been completely understood and illiteracy problem remains unsolved. To this end, P300 peak has been used as the feature of ERP in most brain–computer interface applications, but subjects who do not show such peak are common. Recent development of convolutional neural network provides a way to analyze spatial and temporal features of ERP. Here, we train the convolutional neural network with 2 convolutional layers whose feature maps represented spatial and temporal features of event-related potential. We have found that nonilliterate subjects' ERP show high correlation between occipital lobe and parietal lobe, whereas illiterate subjects only show correlation between neural activities from frontal lobe and central lobe. The nonilliterates showed peaks in P300, P500, and P700, whereas illiterates mostly showed peaks in around P700. P700 was strong in both subjects. We found that P700 peak may be the key feature of ERP as it appears in both illiterate and nonilliterate subjects.

Deep Visual Attention Prediction

NASA Astrophysics Data System (ADS)

Wang, Wenguan; Shen, Jianbing

2018-05-01

In this work, we aim to predict human eye fixation with view-free scenes based on an end-to-end deep learning architecture. Although Convolutional Neural Networks (CNNs) have made substantial improvement on human attention prediction, it is still needed to improve CNN based attention models by efficiently leveraging multi-scale features. Our visual attention network is proposed to capture hierarchical saliency information from deep, coarse layers with global saliency information to shallow, fine layers with local saliency response. Our model is based on a skip-layer network structure, which predicts human attention from multiple convolutional layers with various reception fields. Final saliency prediction is achieved via the cooperation of those global and local predictions. Our model is learned in a deep supervision manner, where supervision is directly fed into multi-level layers, instead of previous approaches of providing supervision only at the output layer and propagating this supervision back to earlier layers. Our model thus incorporates multi-level saliency predictions within a single network, which significantly decreases the redundancy of previous approaches of learning multiple network streams with different input scales. Extensive experimental analysis on various challenging benchmark datasets demonstrate our method yields state-of-the-art performance with competitive inference time.

Deep neural mapping support vector machines.

PubMed

Li, Yujian; Zhang, Ting

2017-09-01

The choice of kernel has an important effect on the performance of a support vector machine (SVM). The effect could be reduced by NEUROSVM, an architecture using multilayer perceptron for feature extraction and SVM for classification. In binary classification, a general linear kernel NEUROSVM can be theoretically simplified as an input layer, many hidden layers, and an SVM output layer. As a feature extractor, the sub-network composed of the input and hidden layers is first trained together with a virtual ordinary output layer by backpropagation, then with the output of its last hidden layer taken as input of the SVM classifier for further training separately. By taking the sub-network as a kernel mapping from the original input space into a feature space, we present a novel model, called deep neural mapping support vector machine (DNMSVM), from the viewpoint of deep learning. This model is also a new and general kernel learning method, where the kernel mapping is indeed an explicit function expressed as a sub-network, different from an implicit function induced by a kernel function traditionally. Moreover, we exploit a two-stage procedure of contrastive divergence learning and gradient descent for DNMSVM to jointly training an adaptive kernel mapping instead of a kernel function, without requirement of kernel tricks. As a whole of the sub-network and the SVM classifier, the joint training of DNMSVM is done by using gradient descent to optimize the objective function with the sub-network layer-wise pre-trained via contrastive divergence learning of restricted Boltzmann machines. Compared to the separate training of NEUROSVM, the joint training is a new algorithm for DNMSVM to have advantages over NEUROSVM. Experimental results show that DNMSVM can outperform NEUROSVM and RBFSVM (i.e., SVM with the kernel of radial basis function), demonstrating its effectiveness. Copyright © 2017 Elsevier Ltd. All rights reserved.

Brain tumor segmentation in multi-spectral MRI using convolutional neural networks (CNN).

PubMed

Iqbal, Sajid; Ghani, M Usman; Saba, Tanzila; Rehman, Amjad

2018-04-01

A tumor could be found in any area of the brain and could be of any size, shape, and contrast. There may exist multiple tumors of different types in a human brain at the same time. Accurate tumor area segmentation is considered primary step for treatment of brain tumors. Deep Learning is a set of promising techniques that could provide better results as compared to nondeep learning techniques for segmenting timorous part inside a brain. This article presents a deep convolutional neural network (CNN) to segment brain tumors in MRIs. The proposed network uses BRATS segmentation challenge dataset which is composed of images obtained through four different modalities. Accordingly, we present an extended version of existing network to solve segmentation problem. The network architecture consists of multiple neural network layers connected in sequential order with the feeding of Convolutional feature maps at the peer level. Experimental results on BRATS 2015 benchmark data thus show the usability of the proposed approach and its superiority over the other approaches in this area of research. © 2018 Wiley Periodicals, Inc.

Linking macroscopic with microscopic neuroanatomy using synthetic neuronal populations.

PubMed

Schneider, Calvin J; Cuntz, Hermann; Soltesz, Ivan

2014-10-01

Dendritic morphology has been shown to have a dramatic impact on neuronal function. However, population features such as the inherent variability in dendritic morphology between cells belonging to the same neuronal type are often overlooked when studying computation in neural networks. While detailed models for morphology and electrophysiology exist for many types of single neurons, the role of detailed single cell morphology in the population has not been studied quantitatively or computationally. Here we use the structural context of the neural tissue in which dendritic trees exist to drive their generation in silico. We synthesize the entire population of dentate gyrus granule cells, the most numerous cell type in the hippocampus, by growing their dendritic trees within their characteristic dendritic fields bounded by the realistic structural context of (1) the granule cell layer that contains all somata and (2) the molecular layer that contains the dendritic forest. This process enables branching statistics to be linked to larger scale neuroanatomical features. We find large differences in dendritic total length and individual path length measures as a function of location in the dentate gyrus and of somatic depth in the granule cell layer. We also predict the number of unique granule cell dendrites invading a given volume in the molecular layer. This work enables the complete population-level study of morphological properties and provides a framework to develop complex and realistic neural network models.

Linking Macroscopic with Microscopic Neuroanatomy Using Synthetic Neuronal Populations

PubMed Central

Schneider, Calvin J.; Cuntz, Hermann; Soltesz, Ivan

2014-01-01

Dendritic morphology has been shown to have a dramatic impact on neuronal function. However, population features such as the inherent variability in dendritic morphology between cells belonging to the same neuronal type are often overlooked when studying computation in neural networks. While detailed models for morphology and electrophysiology exist for many types of single neurons, the role of detailed single cell morphology in the population has not been studied quantitatively or computationally. Here we use the structural context of the neural tissue in which dendritic trees exist to drive their generation in silico. We synthesize the entire population of dentate gyrus granule cells, the most numerous cell type in the hippocampus, by growing their dendritic trees within their characteristic dendritic fields bounded by the realistic structural context of (1) the granule cell layer that contains all somata and (2) the molecular layer that contains the dendritic forest. This process enables branching statistics to be linked to larger scale neuroanatomical features. We find large differences in dendritic total length and individual path length measures as a function of location in the dentate gyrus and of somatic depth in the granule cell layer. We also predict the number of unique granule cell dendrites invading a given volume in the molecular layer. This work enables the complete population-level study of morphological properties and provides a framework to develop complex and realistic neural network models. PMID:25340814

Pluripotency factors in embryonic stem cells regulate differentiation into germ layers.

PubMed

Thomson, Matt; Liu, Siyuan John; Zou, Ling-Nan; Smith, Zack; Meissner, Alexander; Ramanathan, Sharad

2011-06-10

Cell fate decisions are fundamental for development, but we do not know how transcriptional networks reorganize during the transition from a pluripotent to a differentiated cell state. Here, we asked how mouse embryonic stem cells (ESCs) leave the pluripotent state and choose between germ layer fates. By analyzing the dynamics of the transcriptional circuit that maintains pluripotency, we found that Oct4 and Sox2, proteins that maintain ESC identity, also orchestrate germ layer fate selection. Oct4 suppresses neural ectodermal differentiation and promotes mesendodermal differentiation; Sox2 inhibits mesendodermal differentiation and promotes neural ectodermal differentiation. Differentiation signals continuously and asymmetrically modulate Oct4 and Sox2 protein levels, altering their binding pattern in the genome, and leading to cell fate choice. The same factors that maintain pluripotency thus also integrate external signals and control lineage selection. Our study provides a framework for understanding how complex transcription factor networks control cell fate decisions in progenitor cells. Copyright © 2011 Elsevier Inc. All rights reserved.

Multi-layer holographic bifurcative neural network system for real-time adaptive EOS data analysis

NASA Technical Reports Server (NTRS)

Liu, Hua-Kuang; Huang, K. S.; Diep, J.

1993-01-01

Optical data processing techniques have the inherent advantage of high data throughout, low weight and low power requirements. These features are particularly desirable for onboard spacecraft in-situ real-time data analysis and data compression applications. the proposed multi-layer optical holographic neural net pattern recognition technique will utilize the nonlinear photorefractive devices for real-time adaptive learning to classify input data content and recognize unexpected features. Information can be stored either in analog or digital form in a nonlinear photofractive device. The recording can be accomplished in time scales ranging from milliseconds to microseconds. When a system consisting of these devices is organized in a multi-layer structure, a feedforward neural net with bifurcating data classification capability is formed. The interdisciplinary research will involve the collaboration with top digital computer architecture experts at the University of Southern California.

Multilayer neural networks for reduced-rank approximation.

PubMed

Diamantaras, K I; Kung, S Y

1994-01-01

This paper is developed in two parts. First, the authors formulate the solution to the general reduced-rank linear approximation problem relaxing the invertibility assumption of the input autocorrelation matrix used by previous authors. The authors' treatment unifies linear regression, Wiener filtering, full rank approximation, auto-association networks, SVD and principal component analysis (PCA) as special cases. The authors' analysis also shows that two-layer linear neural networks with reduced number of hidden units, trained with the least-squares error criterion, produce weights that correspond to the generalized singular value decomposition of the input-teacher cross-correlation matrix and the input data matrix. As a corollary the linear two-layer backpropagation model with reduced hidden layer extracts an arbitrary linear combination of the generalized singular vector components. Second, the authors investigate artificial neural network models for the solution of the related generalized eigenvalue problem. By introducing and utilizing the extended concept of deflation (originally proposed for the standard eigenvalue problem) the authors are able to find that a sequential version of linear BP can extract the exact generalized eigenvector components. The advantage of this approach is that it's easier to update the model structure by adding one more unit or pruning one or more units when the application requires it. An alternative approach for extracting the exact components is to use a set of lateral connections among the hidden units trained in such a way as to enforce orthogonality among the upper- and lower-layer weights. The authors call this the lateral orthogonalization network (LON) and show via theoretical analysis-and verify via simulation-that the network extracts the desired components. The advantage of the LON-based model is that it can be applied in a parallel fashion so that the components are extracted concurrently. Finally, the authors show the application of their results to the solution of the identification problem of systems whose excitation has a non-invertible autocorrelation matrix. Previous identification methods usually rely on the invertibility assumption of the input autocorrelation, therefore they can not be applied to this case.

Forecasting the prognosis of choroidal melanoma with an artificial neural network.

PubMed

Kaiserman, Igor; Rosner, Mordechai; Pe'er, Jacob

2005-09-01

To develop an artificial neural network (ANN) that will forecast the 5-year mortality from choroidal melanoma. Retrospective, comparative, observational cohort study. One hundred fifty-three eyes of 153 consecutive patients with choroidal melanoma (age, 58.4+/-14.6 years) who were treated with ruthenium 106 brachytherapy between 1988 and 1998 at the Department of Ophthalmology, Hadassah University Hospital, Jerusalem, Israel. Patients were observed clinically and ultrasonographically (A- and B-mode standardized ultrasonography). Metastatic screening included liver function tests and liver imaging. Backpropagation ANNs composed of 3 or 4 layers of neurons with various types of transfer functions and training protocols were assessed for their ability to predict the 5-year mortality. The ANNs were trained on 77 randomly selected patients and tested on a different set of 76 patients. Artificial neural networks were compared based on their sensitivity, specificity, forecasting accuracy, area under the receiver operating curves, and likelihood ratios (LRs). The best ANN was compared with the results of logistic regression and the performance of an ocular oncologist. The ability of the ANNs to forecast the 5-year mortality from choroidal melanoma. Thirty-one patients died during the follow-up period of metastatic choroidal melanoma. The best ANN (one hidden layer of 16 neurons) had 84% forecasting accuracy and an LR of 31.5. The number of hidden neurons significantly influenced the ANNs' performance (P<0.001). The performance of the ANNs was not significantly influenced by the training protocol, the number of hidden layers, or the type of transfer function. In comparison, logistic regression reached 86% forecasting accuracy, with a very low LR (0.8), whereas the human expert forecasting ability was <70% (LR, 1.85). Artificial neural networks can be used for forecasting the prognosis of choroidal melanoma and may support decision-making in treating this malignancy.

Convolutional Neural Networks as Feature Extractors for Data Scarce Visual Searches

DTIC Science & Technology

2016-09-01

50 Figure C.16 TP for the Ferrari Logo (layer=FC7). . . . . . . . . . . . . . . . 50 Figure C.17 TP for the Kia Logo (layer=FC6...51 Figure C.18 TP for the Kia Logo (layer=FC7). . . . . . . . . . . . . . . . . . 51 Figure C.19 TP for the Liege Logo (layer=FC6...represent results for test samples from the Ferrari class for k=1, k=3, and k=5. 50 Figure C.17. TP for the Kia Logo (layer=FC6). The x-axis represents

A fully convolutional networks (FCN) based image segmentation algorithm in binocular imaging system

NASA Astrophysics Data System (ADS)

Long, Zourong; Wei, Biao; Feng, Peng; Yu, Pengwei; Liu, Yuanyuan

2018-01-01

This paper proposes an image segmentation algorithm with fully convolutional networks (FCN) in binocular imaging system under various circumstance. Image segmentation is perfectly solved by semantic segmentation. FCN classifies the pixels, so as to achieve the level of image semantic segmentation. Different from the classical convolutional neural networks (CNN), FCN uses convolution layers instead of the fully connected layers. So it can accept image of arbitrary size. In this paper, we combine the convolutional neural network and scale invariant feature matching to solve the problem of visual positioning under different scenarios. All high-resolution images are captured with our calibrated binocular imaging system and several groups of test data are collected to verify this method. The experimental results show that the binocular images are effectively segmented without over-segmentation. With these segmented images, feature matching via SURF method is implemented to obtain regional information for further image processing. The final positioning procedure shows that the results are acceptable in the range of 1.4 1.6 m, the distance error is less than 10mm.

A fresh look at functional link neural network for motor imagery-based brain-computer interface.

PubMed

Hettiarachchi, Imali T; Babaei, Toktam; Nguyen, Thanh; Lim, Chee P; Nahavandi, Saeid

2018-05-04

Artificial neural networks (ANNs) are one of the widely used classifiers in the brain-computer interface (BCI) systems-based on noninvasive electroencephalography (EEG) signals. Among the different ANN architectures, the most commonly applied for BCI classifiers is the multilayer perceptron (MLP). When appropriately designed with optimal number of neuron layers and number of neurons per layer, the ANN can act as a universal approximator. However, due to the low signal-to-noise ratio of EEG signal data, overtraining problem may become an inherent issue, causing these universal approximators to fail in real-time applications. In this study we introduce a higher order neural network, namely the functional link neural network (FLNN) as a classifier for motor imagery (MI)-based BCI systems, to remedy the drawbacks in MLP. We compare the proposed method with competing classifiers such as linear decomposition analysis, naïve Bayes, k-nearest neighbours, support vector machine and three MLP architectures. Two multi-class benchmark datasets from the BCI competitions are used. Common spatial pattern algorithm is utilized for feature extraction to build classification models. FLNN reports the highest average Kappa value over multiple subjects for both the BCI competition datasets, under similarly preprocessed data and extracted features. Further, statistical comparison results over multiple subjects show that the proposed FLNN classification method yields the best performance among the competing classifiers. Findings from this study imply that the proposed method, which has less computational complexity compared to the MLP, can be implemented effectively in practical MI-based BCI systems. Copyright © 2018 Elsevier B.V. All rights reserved.

Probing many-body localization with neural networks

NASA Astrophysics Data System (ADS)

Schindler, Frank; Regnault, Nicolas; Neupert, Titus

2017-06-01

We show that a simple artificial neural network trained on entanglement spectra of individual states of a many-body quantum system can be used to determine the transition between a many-body localized and a thermalizing regime. Specifically, we study the Heisenberg spin-1/2 chain in a random external field. We employ a multilayer perceptron with a single hidden layer, which is trained on labeled entanglement spectra pertaining to the fully localized and fully thermal regimes. We then apply this network to classify spectra belonging to states in the transition region. For training, we use a cost function that contains, in addition to the usual error and regularization parts, a term that favors a confident classification of the transition region states. The resulting phase diagram is in good agreement with the one obtained by more conventional methods and can be computed for small systems. In particular, the neural network outperforms conventional methods in classifying individual eigenstates pertaining to a single disorder realization. It allows us to map out the structure of these eigenstates across the transition with spatial resolution. Furthermore, we analyze the network operation using the dreaming technique to show that the neural network correctly learns by itself the power-law structure of the entanglement spectra in the many-body localized regime.

Optimal source coding, removable noise elimination, and natural coordinate system construction for general vector sources using replicator neural networks

NASA Astrophysics Data System (ADS)

Hecht-Nielsen, Robert

1997-04-01

A new universal one-chart smooth manifold model for vector information sources is introduced. Natural coordinates (a particular type of chart) for such data manifolds are then defined. Uniformly quantized natural coordinates form an optimal vector quantization code for a general vector source. Replicator neural networks (a specialized type of multilayer perceptron with three hidden layers) are the introduced. As properly configured examples of replicator networks approach minimum mean squared error (e.g., via training and architecture adjustment using randomly chosen vectors from the source), these networks automatically develop a mapping which, in the limit, produces natural coordinates for arbitrary source vectors. The new concept of removable noise (a noise model applicable to a wide variety of real-world noise processes) is then discussed. Replicator neural networks, when configured to approach minimum mean squared reconstruction error (e.g., via training and architecture adjustment on randomly chosen examples from a vector source, each with randomly chosen additive removable noise contamination), in the limit eliminate removable noise and produce natural coordinates for the data vector portions of the noise-corrupted source vectors. Consideration regarding selection of the dimension of a data manifold source model and the training/configuration of replicator neural networks are discussed.

Multilayer motif analysis of brain networks

NASA Astrophysics Data System (ADS)

Battiston, Federico; Nicosia, Vincenzo; Chavez, Mario; Latora, Vito

2017-04-01

In the last decade, network science has shed new light both on the structural (anatomical) and on the functional (correlations in the activity) connectivity among the different areas of the human brain. The analysis of brain networks has made possible to detect the central areas of a neural system and to identify its building blocks by looking at overabundant small subgraphs, known as motifs. However, network analysis of the brain has so far mainly focused on anatomical and functional networks as separate entities. The recently developed mathematical framework of multi-layer networks allows us to perform an analysis of the human brain where the structural and functional layers are considered together. In this work, we describe how to classify the subgraphs of a multiplex network, and we extend the motif analysis to networks with an arbitrary number of layers. We then extract multi-layer motifs in brain networks of healthy subjects by considering networks with two layers, anatomical and functional, respectively, obtained from diffusion and functional magnetic resonance imaging. Results indicate that subgraphs in which the presence of a physical connection between brain areas (links at the structural layer) coexists with a non-trivial positive correlation in their activities are statistically overabundant. Finally, we investigate the existence of a reinforcement mechanism between the two layers by looking at how the probability to find a link in one layer depends on the intensity of the connection in the other one. Showing that functional connectivity is non-trivially constrained by the underlying anatomical network, our work contributes to a better understanding of the interplay between the structure and function in the human brain.

Feasibility study of robotic neural controllers

NASA Technical Reports Server (NTRS)

Magana, Mario E.

1990-01-01

The results are given of a feasibility study performed to establish if an artificial neural controller could be used to achieve joint space trajectory tracking of a two-link robot manipulator. The study is based on the results obtained by Hecht-Nielsen, who claims that a functional map can be implemented to a desired degree of accuracy with a three layer feedforward artificial neural network. Central to this study is the assumption that the robot model as well as its parameters values are known.

Do Convolutional Neural Networks Learn Class Hierarchy?

PubMed

Bilal, Alsallakh; Jourabloo, Amin; Ye, Mao; Liu, Xiaoming; Ren, Liu

2018-01-01

Convolutional Neural Networks (CNNs) currently achieve state-of-the-art accuracy in image classification. With a growing number of classes, the accuracy usually drops as the possibilities of confusion increase. Interestingly, the class confusion patterns follow a hierarchical structure over the classes. We present visual-analytics methods to reveal and analyze this hierarchy of similar classes in relation with CNN-internal data. We found that this hierarchy not only dictates the confusion patterns between the classes, it furthermore dictates the learning behavior of CNNs. In particular, the early layers in these networks develop feature detectors that can separate high-level groups of classes quite well, even after a few training epochs. In contrast, the latter layers require substantially more epochs to develop specialized feature detectors that can separate individual classes. We demonstrate how these insights are key to significant improvement in accuracy by designing hierarchy-aware CNNs that accelerate model convergence and alleviate overfitting. We further demonstrate how our methods help in identifying various quality issues in the training data.

Analyzing Distributional Learning of Phonemic Categories in Unsupervised Deep Neural Networks

PubMed Central

Räsänen, Okko; Nagamine, Tasha; Mesgarani, Nima

2017-01-01

Infants’ speech perception adapts to the phonemic categories of their native language, a process assumed to be driven by the distributional properties of speech. This study investigates whether deep neural networks (DNNs), the current state-of-the-art in distributional feature learning, are capable of learning phoneme-like representations of speech in an unsupervised manner. We trained DNNs with unlabeled and labeled speech and analyzed the activations of each layer with respect to the phones in the input segments. The analyses reveal that the emergence of phonemic invariance in DNNs is dependent on the availability of phonemic labeling of the input during the training. No increased phonemic selectivity of the hidden layers was observed in the purely unsupervised networks despite successful learning of low-dimensional representations for speech. This suggests that additional learning constraints or more sophisticated models are needed to account for the emergence of phone-like categories in distributional learning operating on natural speech. PMID:29359204

Matching algorithm of missile tail flame based on back-propagation neural network

NASA Astrophysics Data System (ADS)

Huang, Da; Huang, Shucai; Tang, Yidong; Zhao, Wei; Cao, Wenhuan

2018-02-01

This work presents a spectral matching algorithm of missile plume detection that based on neural network. The radiation value of the characteristic spectrum of the missile tail flame is taken as the input of the network. The network's structure including the number of nodes and layers is determined according to the number of characteristic spectral bands and missile types. We can get the network weight matrixes and threshold vectors through training the network using training samples, and we can determine the performance of the network through testing the network using the test samples. A small amount of data cause the network has the advantages of simple structure and practicality. Network structure composed of weight matrix and threshold vector can complete task of spectrum matching without large database support. Network can achieve real-time requirements with a small quantity of data. Experiment results show that the algorithm has the ability to match the precise spectrum and strong robustness.

A neural network based computational model to predict the output power of different types of photovoltaic cells.

PubMed

Xiao, WenBo; Nazario, Gina; Wu, HuaMing; Zhang, HuaMing; Cheng, Feng

2017-01-01

In this article, we introduced an artificial neural network (ANN) based computational model to predict the output power of three types of photovoltaic cells, mono-crystalline (mono-), multi-crystalline (multi-), and amorphous (amor-) crystalline. The prediction results are very close to the experimental data, and were also influenced by numbers of hidden neurons. The order of the solar generation power output influenced by the external conditions from smallest to biggest is: multi-, mono-, and amor- crystalline silicon cells. In addition, the dependences of power prediction on the number of hidden neurons were studied. For multi- and amorphous crystalline cell, three or four hidden layer units resulted in the high correlation coefficient and low MSEs. For mono-crystalline cell, the best results were achieved at the hidden layer unit of 8.

Neural network modelling of thermal stratification in a solar DHW storage

DOE Office of Scientific and Technical Information (OSTI.GOV)

Geczy-Vig, P.; Farkas, I.

2010-05-15

In this study an artificial neural network (ANN) model is introduced for modelling the layer temperatures in a storage tank of a solar thermal system. The model is based on the measured data of a domestic hot water system. The temperatures distribution in the storage tank divided in 8 equal parts in vertical direction were calculated every 5 min using the average 5 min data of solar radiation, ambient temperature, mass flow rate of collector loop, load and the temperature of the layers in previous time steps. The introduced ANN model consists of two parts describing the load periods andmore » the periods between the loads. The identified model gives acceptable results inside the training interval as the average deviation was 0.22 C during the training and 0.24 C during the validation. (author)« less

Optimization of Artificial Neural Network using Evolutionary Programming for Prediction of Cascading Collapse Occurrence due to the Hidden Failure Effect

NASA Astrophysics Data System (ADS)

Idris, N. H.; Salim, N. A.; Othman, M. M.; Yasin, Z. M.

2018-03-01

This paper presents the Evolutionary Programming (EP) which proposed to optimize the training parameters for Artificial Neural Network (ANN) in predicting cascading collapse occurrence due to the effect of protection system hidden failure. The data has been collected from the probability of hidden failure model simulation from the historical data. The training parameters of multilayer-feedforward with backpropagation has been optimized with objective function to minimize the Mean Square Error (MSE). The optimal training parameters consists of the momentum rate, learning rate and number of neurons in first hidden layer and second hidden layer is selected in EP-ANN. The IEEE 14 bus system has been tested as a case study to validate the propose technique. The results show the reliable prediction of performance validated through MSE and Correlation Coefficient (R).

Elbow joint angle and elbow movement velocity estimation using NARX-multiple layer perceptron neural network model with surface EMG time domain parameters.

PubMed

Raj, Retheep; Sivanandan, K S

2017-01-01

Estimation of elbow dynamics has been the object of numerous investigations. In this work a solution is proposed for estimating elbow movement velocity and elbow joint angle from Surface Electromyography (SEMG) signals. Here the Surface Electromyography signals are acquired from the biceps brachii muscle of human hand. Two time-domain parameters, Integrated EMG (IEMG) and Zero Crossing (ZC), are extracted from the Surface Electromyography signal. The relationship between the time domain parameters, IEMG and ZC with elbow angular displacement and elbow angular velocity during extension and flexion of the elbow are studied. A multiple input-multiple output model is derived for identifying the kinematics of elbow. A Nonlinear Auto Regressive with eXogenous inputs (NARX) structure based multiple layer perceptron neural network (MLPNN) model is proposed for the estimation of elbow joint angle and elbow angular velocity. The proposed NARX MLPNN model is trained using Levenberg-marquardt based algorithm. The proposed model is estimating the elbow joint angle and elbow movement angular velocity with appreciable accuracy. The model is validated using regression coefficient value (R). The average regression coefficient value (R) obtained for elbow angular displacement prediction is 0.9641 and for the elbow anglular velocity prediction is 0.9347. The Nonlinear Auto Regressive with eXogenous inputs (NARX) structure based multiple layer perceptron neural networks (MLPNN) model can be used for the estimation of angular displacement and movement angular velocity of the elbow with good accuracy.

Models of Acetylcholine and Dopamine Signals Differentially Improve Neural Representations

PubMed Central

Holca-Lamarre, Raphaël; Lücke, Jörg; Obermayer, Klaus

2017-01-01

Biological and artificial neural networks (ANNs) represent input signals as patterns of neural activity. In biology, neuromodulators can trigger important reorganizations of these neural representations. For instance, pairing a stimulus with the release of either acetylcholine (ACh) or dopamine (DA) evokes long lasting increases in the responses of neurons to the paired stimulus. The functional roles of ACh and DA in rearranging representations remain largely unknown. Here, we address this question using a Hebbian-learning neural network model. Our aim is both to gain a functional understanding of ACh and DA transmission in shaping biological representations and to explore neuromodulator-inspired learning rules for ANNs. We model the effects of ACh and DA on synaptic plasticity and confirm that stimuli coinciding with greater neuromodulator activation are over represented in the network. We then simulate the physiological release schedules of ACh and DA. We measure the impact of neuromodulator release on the network's representation and on its performance on a classification task. We find that ACh and DA trigger distinct changes in neural representations that both improve performance. The putative ACh signal redistributes neural preferences so that more neurons encode stimulus classes that are challenging for the network. The putative DA signal adapts synaptic weights so that they better match the classes of the task at hand. Our model thus offers a functional explanation for the effects of ACh and DA on cortical representations. Additionally, our learning algorithm yields performances comparable to those of state-of-the-art optimisation methods in multi-layer perceptrons while requiring weaker supervision signals and interacting with synaptically-local weight updates. PMID:28690509

Identifying Corresponding Patches in SAR and Optical Images With a Pseudo-Siamese CNN

NASA Astrophysics Data System (ADS)

Hughes, Lloyd H.; Schmitt, Michael; Mou, Lichao; Wang, Yuanyuan; Zhu, Xiao Xiang

2018-05-01

In this letter, we propose a pseudo-siamese convolutional neural network (CNN) architecture that enables to solve the task of identifying corresponding patches in very-high-resolution (VHR) optical and synthetic aperture radar (SAR) remote sensing imagery. Using eight convolutional layers each in two parallel network streams, a fully connected layer for the fusion of the features learned in each stream, and a loss function based on binary cross-entropy, we achieve a one-hot indication if two patches correspond or not. The network is trained and tested on an automatically generated dataset that is based on a deterministic alignment of SAR and optical imagery via previously reconstructed and subsequently co-registered 3D point clouds. The satellite images, from which the patches comprising our dataset are extracted, show a complex urban scene containing many elevated objects (i.e. buildings), thus providing one of the most difficult experimental environments. The achieved results show that the network is able to predict corresponding patches with high accuracy, thus indicating great potential for further development towards a generalized multi-sensor key-point matching procedure. Index Terms-synthetic aperture radar (SAR), optical imagery, data fusion, deep learning, convolutional neural networks (CNN), image matching, deep matching

Biological engineering applications of feedforward neural networks designed and parameterized by genetic algorithms.

PubMed

Ferentinos, Konstantinos P

2005-09-01

Two neural network (NN) applications in the field of biological engineering are developed, designed and parameterized by an evolutionary method based on the evolutionary process of genetic algorithms. The developed systems are a fault detection NN model and a predictive modeling NN system. An indirect or 'weak specification' representation was used for the encoding of NN topologies and training parameters into genes of the genetic algorithm (GA). Some a priori knowledge of the demands in network topology for specific application cases is required by this approach, so that the infinite search space of the problem is limited to some reasonable degree. Both one-hidden-layer and two-hidden-layer network architectures were explored by the GA. Except for the network architecture, each gene of the GA also encoded the type of activation functions in both hidden and output nodes of the NN and the type of minimization algorithm that was used by the backpropagation algorithm for the training of the NN. Both models achieved satisfactory performance, while the GA system proved to be a powerful tool that can successfully replace the problematic trial-and-error approach that is usually used for these tasks.

Correlations Decrease with Propagation of Spiking Activity in the Mouse Barrel Cortex

PubMed Central

Ranganathan, Gayathri Nattar; Koester, Helmut Joachim

2011-01-01

Propagation of suprathreshold spiking activity through neuronal populations is important for the function of the central nervous system. Neural correlations have an impact on cortical function particularly on the signaling of information and propagation of spiking activity. Therefore we measured the change in correlations as suprathreshold spiking activity propagated between recurrent neuronal networks of the mammalian cerebral cortex. Using optical methods we recorded spiking activity from large samples of neurons from two neural populations simultaneously. The results indicate that correlations decreased as spiking activity propagated from layer 4 to layer 2/3 in the rodent barrel cortex. PMID:21629764

Neural Networks

DOE Office of Scientific and Technical Information (OSTI.GOV)

Smith, Patrick I.

2003-09-23

Physicists use large detectors to measure particles created in high-energy collisions at particle accelerators. These detectors typically produce signals indicating either where ionization occurs along the path of the particle, or where energy is deposited by the particle. The data produced by these signals is fed into pattern recognition programs to try to identify what particles were produced, and to measure the energy and direction of these particles. Ideally, there are many techniques used in this pattern recognition software. One technique, neural networks, is particularly suitable for identifying what type of particle caused by a set of energy deposits. Neuralmore » networks can derive meaning from complicated or imprecise data, extract patterns, and detect trends that are too complex to be noticed by either humans or other computer related processes. To assist in the advancement of this technology, Physicists use a tool kit to experiment with several neural network techniques. The goal of this research is interface a neural network tool kit into Java Analysis Studio (JAS3), an application that allows data to be analyzed from any experiment. As the final result, a physicist will have the ability to train, test, and implement a neural network with the desired output while using JAS3 to analyze the results or output. Before an implementation of a neural network can take place, a firm understanding of what a neural network is and how it works is beneficial. A neural network is an artificial representation of the human brain that tries to simulate the learning process [5]. It is also important to think of the word artificial in that definition as computer programs that use calculations during the learning process. In short, a neural network learns by representative examples. Perhaps the easiest way to describe the way neural networks learn is to explain how the human brain functions. The human brain contains billions of neural cells that are responsible for processing information [2]. Each one of these cells acts as a simple processor. When individual cells interact with one another, the complex abilities of the brain are made possible. In neural networks, the input or data are processed by a propagation function that adds up the values of all the incoming data. The ending value is then compared with a threshold or specific value. The resulting value must exceed the activation function value in order to become output. The activation function is a mathematical function that a neuron uses to produce an output referring to its input value. [8] Figure 1 depicts this process. Neural networks usually have three components an input, a hidden, and an output. These layers create the end result of the neural network. A real world example is a child associating the word dog with a picture. The child says dog and simultaneously looks a picture of a dog. The input is the spoken word ''dog'', the hidden is the brain processing, and the output will be the category of the word dog based on the picture. This illustration describes how a neural network functions.« less

Information flow in layered networks of non-monotonic units

NASA Astrophysics Data System (ADS)

Schittler Neves, Fabio; Martim Schubert, Benno; Erichsen, Rubem, Jr.

2015-07-01

Layered neural networks are feedforward structures that yield robust parallel and distributed pattern recognition. Even though much attention has been paid to pattern retrieval properties in such systems, many aspects of their dynamics are not yet well characterized or understood. In this work we study, at different temperatures, the memory activity and information flows through layered networks in which the elements are the simplest binary odd non-monotonic function. Our results show that, considering a standard Hebbian learning approach, the network information content has its maximum always at the monotonic limit, even though the maximum memory capacity can be found at non-monotonic values for small enough temperatures. Furthermore, we show that such systems exhibit rich macroscopic dynamics, including not only fixed point solutions of its iterative map, but also cyclic and chaotic attractors that also carry information.

Regional application of multi-layer artificial neural networks in 3-D ionosphere tomography

NASA Astrophysics Data System (ADS)

Ghaffari Razin, Mir Reza; Voosoghi, Behzad

2016-08-01

Tomography is a very cost-effective method to study physical properties of the ionosphere. In this paper, residual minimization training neural network (RMTNN) is used in voxel-based tomography to reconstruct of 3-D ionosphere electron density with high spatial resolution. For numerical experiments, observations collected at 37 GPS stations from Iranian permanent GPS network (IPGN) are used. A smoothed TEC approach was used for absolute STEC recovery. To improve the vertical resolution, empirical orthogonal functions (EOFs) obtained from international reference ionosphere 2012 (IRI-2012) used as object function in training neural network. Ionosonde observations is used for validate reliability of the proposed method. Minimum relative error for RMTNN is 1.64% and maximum relative error is 15.61%. Also root mean square error (RMSE) of 0.17 × 1011 (electrons/m3) is computed for RMTNN which is less than RMSE of IRI2012. The results show that RMTNN has higher accuracy and compiles speed than other ionosphere reconstruction methods.

Detection of high-grade small bowel obstruction on conventional radiography with convolutional neural networks.

PubMed

Cheng, Phillip M; Tejura, Tapas K; Tran, Khoa N; Whang, Gilbert

2018-05-01

The purpose of this pilot study is to determine whether a deep convolutional neural network can be trained with limited image data to detect high-grade small bowel obstruction patterns on supine abdominal radiographs. Grayscale images from 3663 clinical supine abdominal radiographs were categorized into obstructive and non-obstructive categories independently by three abdominal radiologists, and the majority classification was used as ground truth; 74 images were found to be consistent with small bowel obstruction. Images were rescaled and randomized, with 2210 images constituting the training set (39 with small bowel obstruction) and 1453 images constituting the test set (35 with small bowel obstruction). Weight parameters for the final classification layer of the Inception v3 convolutional neural network, previously trained on the 2014 Large Scale Visual Recognition Challenge dataset, were retrained on the training set. After training, the neural network achieved an AUC of 0.84 on the test set (95% CI 0.78-0.89). At the maximum Youden index (sensitivity + specificity-1), the sensitivity of the system for small bowel obstruction is 83.8%, with a specificity of 68.1%. The results demonstrate that transfer learning with convolutional neural networks, even with limited training data, may be used to train a detector for high-grade small bowel obstruction gas patterns on supine radiographs.

Comparison of Logistic Regression and Artificial Neural Network in Low Back Pain Prediction: Second National Health Survey

PubMed Central

Parsaeian, M; Mohammad, K; Mahmoudi, M; Zeraati, H

2012-01-01

Background: The purpose of this investigation was to compare empirically predictive ability of an artificial neural network with a logistic regression in prediction of low back pain. Methods: Data from the second national health survey were considered in this investigation. This data includes the information of low back pain and its associated risk factors among Iranian people aged 15 years and older. Artificial neural network and logistic regression models were developed using a set of 17294 data and they were validated in a test set of 17295 data. Hosmer and Lemeshow recommendation for model selection was used in fitting the logistic regression. A three-layer perceptron with 9 inputs, 3 hidden and 1 output neurons was employed. The efficiency of two models was compared by receiver operating characteristic analysis, root mean square and -2 Loglikelihood criteria. Results: The area under the ROC curve (SE), root mean square and -2Loglikelihood of the logistic regression was 0.752 (0.004), 0.3832 and 14769.2, respectively. The area under the ROC curve (SE), root mean square and -2Loglikelihood of the artificial neural network was 0.754 (0.004), 0.3770 and 14757.6, respectively. Conclusions: Based on these three criteria, artificial neural network would give better performance than logistic regression. Although, the difference is statistically significant, it does not seem to be clinically significant. PMID:23113198

Comparison of logistic regression and artificial neural network in low back pain prediction: second national health survey.

PubMed

Parsaeian, M; Mohammad, K; Mahmoudi, M; Zeraati, H

2012-01-01

The purpose of this investigation was to compare empirically predictive ability of an artificial neural network with a logistic regression in prediction of low back pain. Data from the second national health survey were considered in this investigation. This data includes the information of low back pain and its associated risk factors among Iranian people aged 15 years and older. Artificial neural network and logistic regression models were developed using a set of 17294 data and they were validated in a test set of 17295 data. Hosmer and Lemeshow recommendation for model selection was used in fitting the logistic regression. A three-layer perceptron with 9 inputs, 3 hidden and 1 output neurons was employed. The efficiency of two models was compared by receiver operating characteristic analysis, root mean square and -2 Loglikelihood criteria. The area under the ROC curve (SE), root mean square and -2Loglikelihood of the logistic regression was 0.752 (0.004), 0.3832 and 14769.2, respectively. The area under the ROC curve (SE), root mean square and -2Loglikelihood of the artificial neural network was 0.754 (0.004), 0.3770 and 14757.6, respectively. Based on these three criteria, artificial neural network would give better performance than logistic regression. Although, the difference is statistically significant, it does not seem to be clinically significant.

Multi-objective evolutionary optimization for constructing neural networks for virtual reality visual data mining: application to geophysical prospecting.

PubMed

Valdés, Julio J; Barton, Alan J

2007-05-01

A method for the construction of virtual reality spaces for visual data mining using multi-objective optimization with genetic algorithms on nonlinear discriminant (NDA) neural networks is presented. Two neural network layers (the output and the last hidden) are used for the construction of simultaneous solutions for: (i) a supervised classification of data patterns and (ii) an unsupervised similarity structure preservation between the original data matrix and its image in the new space. A set of spaces are constructed from selected solutions along the Pareto front. This strategy represents a conceptual improvement over spaces computed by single-objective optimization. In addition, genetic programming (in particular gene expression programming) is used for finding analytic representations of the complex mappings generating the spaces (a composition of NDA and orthogonal principal components). The presented approach is domain independent and is illustrated via application to the geophysical prospecting of caves.

Deep convolutional neural network for the automated detection and diagnosis of seizure using EEG signals.

PubMed

Acharya, U Rajendra; Oh, Shu Lih; Hagiwara, Yuki; Tan, Jen Hong; Adeli, Hojjat

2017-09-27

An encephalogram (EEG) is a commonly used ancillary test to aide in the diagnosis of epilepsy. The EEG signal contains information about the electrical activity of the brain. Traditionally, neurologists employ direct visual inspection to identify epileptiform abnormalities. This technique can be time-consuming, limited by technical artifact, provides variable results secondary to reader expertise level, and is limited in identifying abnormalities. Therefore, it is essential to develop a computer-aided diagnosis (CAD) system to automatically distinguish the class of these EEG signals using machine learning techniques. This is the first study to employ the convolutional neural network (CNN) for analysis of EEG signals. In this work, a 13-layer deep convolutional neural network (CNN) algorithm is implemented to detect normal, preictal, and seizure classes. The proposed technique achieved an accuracy, specificity, and sensitivity of 88.67%, 90.00% and 95.00%, respectively. Copyright © 2017 Elsevier Ltd. All rights reserved.

Relabeling exchange method (REM) for learning in neural networks

NASA Astrophysics Data System (ADS)

Wu, Wen; Mammone, Richard J.

1994-02-01

The supervised training of neural networks require the use of output labels which are usually arbitrarily assigned. In this paper it is shown that there is a significant difference in the rms error of learning when `optimal' label assignment schemes are used. We have investigated two efficient random search algorithms to solve the relabeling problem: the simulated annealing and the genetic algorithm. However, we found them to be computationally expensive. Therefore we shall introduce a new heuristic algorithm called the Relabeling Exchange Method (REM) which is computationally more attractive and produces optimal performance. REM has been used to organize the optimal structure for multi-layered perceptrons and neural tree networks. The method is a general one and can be implemented as a modification to standard training algorithms. The motivation of the new relabeling strategy is based on the present interpretation of dyslexia as an encoding problem.

Constitutive flow behaviour of austenitic stainless steels under hot deformation: artificial neural network modelling to understand, evaluate and predict

NASA Astrophysics Data System (ADS)

Mandal, Sumantra; Sivaprasad, P. V.; Venugopal, S.; Murthy, K. P. N.

2006-09-01

An artificial neural network (ANN) model is developed to predict the constitutive flow behaviour of austenitic stainless steels during hot deformation. The input parameters are alloy composition and process variables whereas flow stress is the output. The model is based on a three-layer feed-forward ANN with a back-propagation learning algorithm. The neural network is trained with an in-house database obtained from hot compression tests on various grades of austenitic stainless steels. The performance of the model is evaluated using a wide variety of statistical indices. Good agreement between experimental and predicted data is obtained. The correlation between individual alloying elements and high temperature flow behaviour is investigated by employing the ANN model. The results are found to be consistent with the physical phenomena. The model can be used as a guideline for new alloy development.

Baseline estimation in flame's spectra by using neural networks and robust statistics

NASA Astrophysics Data System (ADS)

Garces, Hugo; Arias, Luis; Rojas, Alejandro

2014-09-01

This work presents a baseline estimation method in flame spectra based on artificial intelligence structure as a neural network, combining robust statistics with multivariate analysis to automatically discriminate measured wavelengths belonging to continuous feature for model adaptation, surpassing restriction of measuring target baseline for training. The main contributions of this paper are: to analyze a flame spectra database computing Jolliffe statistics from Principal Components Analysis detecting wavelengths not correlated with most of the measured data corresponding to baseline; to systematically determine the optimal number of neurons in hidden layers based on Akaike's Final Prediction Error; to estimate baseline in full wavelength range sampling measured spectra; and to train an artificial intelligence structure as a Neural Network which allows to generalize the relation between measured and baseline spectra. The main application of our research is to compute total radiation with baseline information, allowing to diagnose combustion process state for optimization in early stages.

Approximating quantum many-body wave functions using artificial neural networks

NASA Astrophysics Data System (ADS)

Cai, Zi; Liu, Jinguo

2018-01-01

In this paper, we demonstrate the expressibility of artificial neural networks (ANNs) in quantum many-body physics by showing that a feed-forward neural network with a small number of hidden layers can be trained to approximate with high precision the ground states of some notable quantum many-body systems. We consider the one-dimensional free bosons and fermions, spinless fermions on a square lattice away from half-filling, as well as frustrated quantum magnetism with a rapidly oscillating ground-state characteristic function. In the latter case, an ANN with a standard architecture fails, while that with a slightly modified one successfully learns the frustration-induced complex sign rule in the ground state and approximates the ground states with high precisions. As an example of practical use of our method, we also perform the variational method to explore the ground state of an antiferromagnetic J1-J2 Heisenberg model.

Neural network hardware and software solutions for sorting of waste plastics for recycling

DOE Office of Scientific and Technical Information (OSTI.GOV)

Stanton, S.L.; Alam, M.K.; Hebner, G.A.

1992-12-31

While plastic recycling efforts have expanded during the past several years, the cost of recovering plastics is still a major impediment for recyclers. Several factors contribute to the prohibitive cost of recycled resins, including the present low marketability of products made with mixed recycled materials, and costs of collecting, sorting and reprocessing plastic materials. A method for automatic sorting of post-consumer plastics into pure polymer streams is needed to overcome the inaccuracies and low product throughput of the currently used method of hand sorting of waste plastics for recycling. The Society of Plastics has designated seven categories as recyclable: Polyethylenemore » terephthalate (PET); High Density Polyethylene (HDPE); Polyvinyl Chloride (PVC); Low Density Polyethylene (LDPE); Polypropylene (PP); Polystyrene (PS); and Other (mixtures, layered items, etc.). With these categories in mind, a system for sorting of waste plastics using near-infrared reflectance spectra and a backpropagation neural network classifier has been developed. A solution has been demonstrated in the laboratory using a high resolution, and relatively slow instrument. A faster instrument is being developed at this time. Neural network hardware options have been evaluated for use in a real-time industrial system. In the lab, a Fourier transform Near Infrared (FT-NIR) scanning spectrometer was used to gather reflectance data from various locations on samples of actual waste plastics. Neural networks were trained off-line with this data using the NeuralWorks Professional II Plus software package on a SparcStation 2. One of the successfully trained networks was used to compare the neural accelerator hardware options available. The results of running this ``worst case`` network on the neural network hardware will be presented. The AT&T ANNA chip and the Intel 80170NX chip development system were used to determine the ease of implementation and accuracies for this network.« less

Neural network hardware and software solutions for sorting of waste plastics for recycling

DOE Office of Scientific and Technical Information (OSTI.GOV)

Stanton, S.L.; Alam, M.K.; Hebner, G.A.

1992-01-01

While plastic recycling efforts have expanded during the past several years, the cost of recovering plastics is still a major impediment for recyclers. Several factors contribute to the prohibitive cost of recycled resins, including the present low marketability of products made with mixed recycled materials, and costs of collecting, sorting and reprocessing plastic materials. A method for automatic sorting of post-consumer plastics into pure polymer streams is needed to overcome the inaccuracies and low product throughput of the currently used method of hand sorting of waste plastics for recycling. The Society of Plastics has designated seven categories as recyclable: Polyethylenemore » terephthalate (PET); High Density Polyethylene (HDPE); Polyvinyl Chloride (PVC); Low Density Polyethylene (LDPE); Polypropylene (PP); Polystyrene (PS); and Other (mixtures, layered items, etc.). With these categories in mind, a system for sorting of waste plastics using near-infrared reflectance spectra and a backpropagation neural network classifier has been developed. A solution has been demonstrated in the laboratory using a high resolution, and relatively slow instrument. A faster instrument is being developed at this time. Neural network hardware options have been evaluated for use in a real-time industrial system. In the lab, a Fourier transform Near Infrared (FT-NIR) scanning spectrometer was used to gather reflectance data from various locations on samples of actual waste plastics. Neural networks were trained off-line with this data using the NeuralWorks Professional II Plus software package on a SparcStation 2. One of the successfully trained networks was used to compare the neural accelerator hardware options available. The results of running this worst case'' network on the neural network hardware will be presented. The AT T ANNA chip and the Intel 80170NX chip development system were used to determine the ease of implementation and accuracies for this network.« less

Mapping the Structure of Directed Networks: Beyond the Bow-Tie Diagram

NASA Astrophysics Data System (ADS)

Timár, G.; Goltsev, A. V.; Dorogovtsev, S. N.; Mendes, J. F. F.

2017-02-01

We reveal a hierarchical, multilayer organization of finite components—i.e., tendrils and tubes—around the giant connected components in directed networks and propose efficient algorithms allowing one to uncover the entire organization of key real-world directed networks, such as the World Wide Web, the neural network of Caenorhabditis elegans, and others. With increasing damage, the giant components decrease in size while the number and size of tendril layers increase, enhancing the susceptibility of the networks to damage.

Comparison of universal approximators incorporating partial monotonicity by structure.

PubMed

Minin, Alexey; Velikova, Marina; Lang, Bernhard; Daniels, Hennie

2010-05-01

Neural networks applied in control loops and safety-critical domains have to meet more requirements than just the overall best function approximation. On the one hand, a small approximation error is required; on the other hand, the smoothness and the monotonicity of selected input-output relations have to be guaranteed. Otherwise, the stability of most of the control laws is lost. In this article we compare two neural network-based approaches incorporating partial monotonicity by structure, namely the Monotonic Multi-Layer Perceptron (MONMLP) network and the Monotonic MIN-MAX (MONMM) network. We show the universal approximation capabilities of both types of network for partially monotone functions. On a number of datasets, we investigate the advantages and disadvantages of these approaches related to approximation performance, training of the model and convergence. 2009 Elsevier Ltd. All rights reserved.

Experimental Verification of Electric Drive Technologies Based on Artificial Intelligence Tools

NASA Technical Reports Server (NTRS)

Rubaai, Ahmed; Ricketts, Daniel; Kotaru, Raj; Thomas, Robert; Noga, Donald F. (Technical Monitor); Kankam, Mark D. (Technical Monitor)

2000-01-01

In this report, a fully integrated prototype of a flight servo control system is successfully developed and implemented using brushless dc motors. The control system is developed by the fuzzy logic theory, and implemented with a multilayer neural network. First, a neural network-based architecture is introduced for fuzzy logic control. The characteristic rules and their membership functions of fuzzy systems are represented as the processing nodes in the neural network structure. The network structure and the parameter learning are performed simultaneously and online in the fuzzy-neural network system. The structure learning is based on the partition of input space. The parameter learning is based on the supervised gradient decent method, using a delta adaptation law. Using experimental setup, the performance of the proposed control system is evaluated under various operating conditions. Test results are presented and discussed in the report. The proposed learning control system has several advantages, namely, simple structure and learning capability, robustness and high tracking performance and few nodes at hidden layers. In comparison with the PI controller, the proposed fuzzy-neural network system can yield a better dynamic performance with shorter settling time, and without overshoot. Experimental results have shown that the proposed control system is adaptive and robust in responding to a wide range of operating conditions. In summary, the goal of this study is to design and implement-advanced servosystems to actuate control surfaces for flight vehicles, namely, aircraft and helicopters, missiles and interceptors, and mini- and micro-air vehicles.

The 3-D image recognition based on fuzzy neural network technology

NASA Technical Reports Server (NTRS)

Hirota, Kaoru; Yamauchi, Kenichi; Murakami, Jun; Tanaka, Kei

1993-01-01

Three dimensional stereoscopic image recognition system based on fuzzy-neural network technology was developed. The system consists of three parts; preprocessing part, feature extraction part, and matching part. Two CCD color camera image are fed to the preprocessing part, where several operations including RGB-HSV transformation are done. A multi-layer perception is used for the line detection in the feature extraction part. Then fuzzy matching technique is introduced in the matching part. The system is realized on SUN spark station and special image input hardware system. An experimental result on bottle images is also presented.

The effect of an exogenous magnetic field on neural coding in deep spiking neural networks.

PubMed

Guo, Lei; Zhang, Wei; Zhang, Jialei

2018-01-01

A ten-layer feed forward network is constructed in the presence of an exogenous alternating magnetic field. Specifically, our results indicate that for rate coding, the firing rate is significantly increased in the presence of an exogenous alternating magnetic field and particularly with increasing enhancement of the alternating magnetic field amplitude. For temporal coding, the interspike intervals of the spiking sequence are decreased and the distribution of the interspike intervals of the spiking sequence tends to be uniform in the presence of alternating magnetic field.

Artificial neural networks and approximate reasoning for intelligent control in space

NASA Technical Reports Server (NTRS)

Berenji, Hamid R.

1991-01-01

A method is introduced for learning to refine the control rules of approximate reasoning-based controllers. A reinforcement-learning technique is used in conjunction with a multi-layer neural network model of an approximate reasoning-based controller. The model learns by updating its prediction of the physical system's behavior. The model can use the control knowledge of an experienced operator and fine-tune it through the process of learning. Some of the space domains suitable for applications of the model such as rendezvous and docking, camera tracking, and tethered systems control are discussed.

Multi-layer network utilizing rewarded spike time dependent plasticity to learn a foraging task

PubMed Central

2017-01-01

Neural networks with a single plastic layer employing reward modulated spike time dependent plasticity (STDP) are capable of learning simple foraging tasks. Here we demonstrate advanced pattern discrimination and continuous learning in a network of spiking neurons with multiple plastic layers. The network utilized both reward modulated and non-reward modulated STDP and implemented multiple mechanisms for homeostatic regulation of synaptic efficacy, including heterosynaptic plasticity, gain control, output balancing, activity normalization of rewarded STDP and hard limits on synaptic strength. We found that addition of a hidden layer of neurons employing non-rewarded STDP created neurons that responded to the specific combinations of inputs and thus performed basic classification of the input patterns. When combined with a following layer of neurons implementing rewarded STDP, the network was able to learn, despite the absence of labeled training data, discrimination between rewarding patterns and the patterns designated as punishing. Synaptic noise allowed for trial-and-error learning that helped to identify the goal-oriented strategies which were effective in task solving. The study predicts a critical set of properties of the spiking neuronal network with STDP that was sufficient to solve a complex foraging task involving pattern classification and decision making. PMID:28961245

Neural Network and Letter Recognition.

NASA Astrophysics Data System (ADS)

Lee, Hue Yeon

Neural net architectures and learning algorithms that recognize hand written 36 alphanumeric characters are studied. The thin line input patterns written in 32 x 32 binary array are used. The system is comprised of two major components, viz. a preprocessing unit and a Recognition unit. The preprocessing unit in turn consists of three layers of neurons; the U-layer, the V-layer, and the C -layer. The functions of the U-layer is to extract local features by template matching. The correlation between the detected local features are considered. Through correlating neurons in a plane with their neighboring neurons, the V-layer would thicken the on-cells or lines that are groups of on-cells of the previous layer. These two correlations would yield some deformation tolerance and some of the rotational tolerance of the system. The C-layer then compresses data through the 'Gabor' transform. Pattern dependent choice of center and wavelengths of 'Gabor' filters is the cause of shift and scale tolerance of the system. Three different learning schemes had been investigated in the recognition unit, namely; the error back propagation learning with hidden units, a simple perceptron learning, and a competitive learning. Their performances were analyzed and compared. Since sometimes the network fails to distinguish between two letters that are inherently similar, additional ambiguity resolving neural nets are introduced on top of the above main neural net. The two dimensional Fourier transform is used as the preprocessing and the perceptron is used as the recognition unit of the ambiguity resolver. One hundred different person's handwriting sets are collected. Some of these are used as the training sets and the remainders are used as the test sets. The correct recognition rate of the system increases with the number of training sets and eventually saturates at a certain value. Similar recognition rates are obtained for the above three different learning algorithms. The minimum error rate, 4.9% is achieved for alphanumeric sets when 50 sets are trained. With the ambiguity resolver, it is reduced to 2.5%. In case that only numeral sets are trained and tested, 2.0% error rate is achieved. When only alphabet sets are considered, the error rate is reduced to 1.1%.

Neural network approach to the inverse problem of the crack-depth determination from ultrasonic backscattering data

NASA Astrophysics Data System (ADS)

Takadoya, M.; Notake, M.; Kitahara, M.; Achenbach, J. D.; Guo, Q. C.; Peterson, M. L.

A neural network approach has been developed to determine the depth of a surface breaking crack in a steel plate from ultrasonic backscattering data. The network is trained by the use of a feedforward three-layered network together with a back-propagation algorithm for error corrections. Synthetic data are employed for network training. The signal used for crack isonification is a mode converted 45 deg transverse wave. The plate with a surface breaking crack is immersed in water, and the crack is insonified from the opposite uncracked side of the plate. A numerical analysis of the backscattered field is carried out based on the elastic wave theory by the use of the boundary element method. The numerical analysis provides synthetic data for the training of the network. The training data have been calculated for cracks with specific increments of the experimental data which are different from the training data.

Decoding of finger trajectory from ECoG using deep learning.

PubMed

Xie, Ziqian; Schwartz, Odelia; Prasad, Abhishek

2018-06-01

Conventional decoding pipeline for brain-machine interfaces (BMIs) consists of chained different stages of feature extraction, time-frequency analysis and statistical learning models. Each of these stages uses a different algorithm trained in a sequential manner, which makes it difficult to make the whole system adaptive. The goal was to create an adaptive online system with a single objective function and a single learning algorithm so that the whole system can be trained in parallel to increase the decoding performance. Here, we used deep neural networks consisting of convolutional neural networks (CNN) and a special kind of recurrent neural network (RNN) called long short term memory (LSTM) to address these needs. We used electrocorticography (ECoG) data collected by Kubanek et al. The task consisted of individual finger flexions upon a visual cue. Our model combined a hierarchical feature extractor CNN and a RNN that was able to process sequential data and recognize temporal dynamics in the neural data. CNN was used as the feature extractor and LSTM was used as the regression algorithm to capture the temporal dynamics of the signal. We predicted the finger trajectory using ECoG signals and compared results for the least angle regression (LARS), CNN-LSTM, random forest, LSTM model (LSTM_HC, for using hard-coded features) and a decoding pipeline consisting of band-pass filtering, energy extraction, feature selection and linear regression. The results showed that the deep learning models performed better than the commonly used linear model. The deep learning models not only gave smoother and more realistic trajectories but also learned the transition between movement and rest state. This study demonstrated a decoding network for BMI that involved a convolutional and recurrent neural network model. It integrated the feature extraction pipeline into the convolution and pooling layer and used LSTM layer to capture the state transitions. The discussed network eliminated the need to separately train the model at each step in the decoding pipeline. The whole system can be jointly optimized using stochastic gradient descent and is capable of online learning.

Decoding of finger trajectory from ECoG using deep learning

NASA Astrophysics Data System (ADS)

Xie, Ziqian; Schwartz, Odelia; Prasad, Abhishek

2018-06-01

Objective. Conventional decoding pipeline for brain-machine interfaces (BMIs) consists of chained different stages of feature extraction, time-frequency analysis and statistical learning models. Each of these stages uses a different algorithm trained in a sequential manner, which makes it difficult to make the whole system adaptive. The goal was to create an adaptive online system with a single objective function and a single learning algorithm so that the whole system can be trained in parallel to increase the decoding performance. Here, we used deep neural networks consisting of convolutional neural networks (CNN) and a special kind of recurrent neural network (RNN) called long short term memory (LSTM) to address these needs. Approach. We used electrocorticography (ECoG) data collected by Kubanek et al. The task consisted of individual finger flexions upon a visual cue. Our model combined a hierarchical feature extractor CNN and a RNN that was able to process sequential data and recognize temporal dynamics in the neural data. CNN was used as the feature extractor and LSTM was used as the regression algorithm to capture the temporal dynamics of the signal. Main results. We predicted the finger trajectory using ECoG signals and compared results for the least angle regression (LARS), CNN-LSTM, random forest, LSTM model (LSTM_HC, for using hard-coded features) and a decoding pipeline consisting of band-pass filtering, energy extraction, feature selection and linear regression. The results showed that the deep learning models performed better than the commonly used linear model. The deep learning models not only gave smoother and more realistic trajectories but also learned the transition between movement and rest state. Significance. This study demonstrated a decoding network for BMI that involved a convolutional and recurrent neural network model. It integrated the feature extraction pipeline into the convolution and pooling layer and used LSTM layer to capture the state transitions. The discussed network eliminated the need to separately train the model at each step in the decoding pipeline. The whole system can be jointly optimized using stochastic gradient descent and is capable of online learning.

Triphasic spike-timing-dependent plasticity organizes networks to produce robust sequences of neural activity

PubMed Central

Waddington, Amelia; Appleby, Peter A.; De Kamps, Marc; Cohen, Netta

2012-01-01

Synfire chains have long been proposed to generate precisely timed sequences of neural activity. Such activity has been linked to numerous neural functions including sensory encoding, cognitive and motor responses. In particular, it has been argued that synfire chains underlie the precise spatiotemporal firing patterns that control song production in a variety of songbirds. Previous studies have suggested that the development of synfire chains requires either initial sparse connectivity or strong topological constraints, in addition to any synaptic learning rules. Here, we show that this necessity can be removed by using a previously reported but hitherto unconsidered spike-timing-dependent plasticity (STDP) rule and activity-dependent excitability. Under this rule the network develops stable synfire chains that possess a non-trivial, scalable multi-layer structure, in which relative layer sizes appear to follow a universal function. Using computational modeling and a coarse grained random walk model, we demonstrate the role of the STDP rule in growing, molding and stabilizing the chain, and link model parameters to the resulting structure. PMID:23162457

Nuevas tecnicas basadas en redes neuronales para el diseno de filtros de microondas multicapa apantallados

NASA Astrophysics Data System (ADS)

Pascual Garcia, Juan

In this PhD thesis one method of shielded multilayer circuit neural network based analysis has been developed. One of the most successful analysis procedures of these kind of structures is the Integral Equation technique (IE) solved by the Method of Moments (MoM). In order to solve the IE, in the version which uses the media relevant potentials, it is necessary to have a formulation of the Green's functions associated to the mentioned potentials. The main computational burden in the IE resolution lies on the numerical evaluation of the Green's functions. In this work, the circuit analysis has been drastically accelerated thanks to the approximation of the Green's functions by means of neural networks. Once trained, the neural networks substitute the Green's functions in the IE. Two different types of neural networks have been used: the Radial basis function neural networks (RBFNN) and the Chebyshev neural networks. Thanks mainly to two distinct operations the correct approximation of the Green's functions has been possible. On the one hand, a very effective input space division has been developed. On the other hand, the elimination of the singularity makes feasible the approximation of slow variation functions. Two different singularity elimination strategies have been developed. The first one is based on the multiplication by the source-observation points distance (rho). The second one outperforms the first one. It consists of the extraction of two layers of spatial images from the whole summation of images. With regard to the Chebyshev neural networks, the OLS training algorithm has been applied in a novel fashion. This method allows the optimum design in this kind of neural networks. In this way, the performance of these neural networks outperforms greatly the RBFNNs one. In both networks, the time gain reached makes the neural method profitable. The time invested in the input space division and in the neural training is negligible with only few circuit analysis. To show, in a practical way, the ability of the neural based analysis method, two new design procedures have been developed. The first method uses the Genetic Algorithms to optimize an initial filter which does not fulfill the established specifications. A new fitness function, specially well suited to design filters, has been defined in order to assure the correct convergence of the optimization process. This new function measures the fulfillment of the specifications and it also prevents the appearance of the premature convergence problem. The second method is found on the approximation, by means of neural networks, of the relations between the electrical parameters, which defined the circuit response, and the physical dimensions that synthesize the aforementioned parameters. The neural networks trained with these data can be used in the design of many circuits in a given structure. Both methods had been show their ability in the design of practical filters.

Are Imaging and Lesioning Convergent Methods for Assessing Functional Specialisation? Investigations Using an Artificial Neural Network

ERIC Educational Resources Information Center

Thomas, Michael S. C.; Purser, Harry R. M.; Tomlinson, Simon; Mareschal, Denis

2012-01-01

This article presents an investigation of the relationship between lesioning and neuroimaging methods of assessing functional specialisation, using synthetic brain imaging (SBI) and lesioning of a connectionist network of past-tense formation. The model comprised two processing "routes": one was a direct route between layers of input and output…

Microfluidic neurite guidance to study structure-function relationships in topologically-complex population-based neural networks.

PubMed

Honegger, Thibault; Thielen, Moritz I; Feizi, Soheil; Sanjana, Neville E; Voldman, Joel

2016-06-22

The central nervous system is a dense, layered, 3D interconnected network of populations of neurons, and thus recapitulating that complexity for in vitro CNS models requires methods that can create defined topologically-complex neuronal networks. Several three-dimensional patterning approaches have been developed but none have demonstrated the ability to control the connections between populations of neurons. Here we report a method using AC electrokinetic forces that can guide, accelerate, slow down and push up neurites in un-modified collagen scaffolds. We present a means to create in vitro neural networks of arbitrary complexity by using such forces to create 3D intersections of primary neuronal populations that are plated in a 2D plane. We report for the first time in vitro basic brain motifs that have been previously observed in vivo and show that their functional network is highly decorrelated to their structure. This platform can provide building blocks to reproduce in vitro the complexity of neural circuits and provide a minimalistic environment to study the structure-function relationship of the brain circuitry.

Microfluidic neurite guidance to study structure-function relationships in topologically-complex population-based neural networks

NASA Astrophysics Data System (ADS)

Honegger, Thibault; Thielen, Moritz I.; Feizi, Soheil; Sanjana, Neville E.; Voldman, Joel

2016-06-01

The central nervous system is a dense, layered, 3D interconnected network of populations of neurons, and thus recapitulating that complexity for in vitro CNS models requires methods that can create defined topologically-complex neuronal networks. Several three-dimensional patterning approaches have been developed but none have demonstrated the ability to control the connections between populations of neurons. Here we report a method using AC electrokinetic forces that can guide, accelerate, slow down and push up neurites in un-modified collagen scaffolds. We present a means to create in vitro neural networks of arbitrary complexity by using such forces to create 3D intersections of primary neuronal populations that are plated in a 2D plane. We report for the first time in vitro basic brain motifs that have been previously observed in vivo and show that their functional network is highly decorrelated to their structure. This platform can provide building blocks to reproduce in vitro the complexity of neural circuits and provide a minimalistic environment to study the structure-function relationship of the brain circuitry.

Visual Saliency Detection Based on Multiscale Deep CNN Features.

PubMed

Guanbin Li; Yizhou Yu

2016-11-01

Visual saliency is a fundamental problem in both cognitive and computational sciences, including computer vision. In this paper, we discover that a high-quality visual saliency model can be learned from multiscale features extracted using deep convolutional neural networks (CNNs), which have had many successes in visual recognition tasks. For learning such saliency models, we introduce a neural network architecture, which has fully connected layers on top of CNNs responsible for feature extraction at three different scales. The penultimate layer of our neural network has been confirmed to be a discriminative high-level feature vector for saliency detection, which we call deep contrast feature. To generate a more robust feature, we integrate handcrafted low-level features with our deep contrast feature. To promote further research and evaluation of visual saliency models, we also construct a new large database of 4447 challenging images and their pixelwise saliency annotations. Experimental results demonstrate that our proposed method is capable of achieving the state-of-the-art performance on all public benchmarks, improving the F-measure by 6.12% and 10%, respectively, on the DUT-OMRON data set and our new data set (HKU-IS), and lowering the mean absolute error by 9% and 35.3%, respectively, on these two data sets.

Modeling the Effects of Cu Content and Deformation Variables on the High-Temperature Flow Behavior of Dilute Al-Fe-Si Alloys Using an Artificial Neural Network.

PubMed

Shakiba, Mohammad; Parson, Nick; Chen, X-Grant

2016-06-30

The hot deformation behavior of Al-0.12Fe-0.1Si alloys with varied amounts of Cu (0.002-0.31 wt %) was investigated by uniaxial compression tests conducted at different temperatures (400 °C-550 °C) and strain rates (0.01-10 s -1 ). The results demonstrated that flow stress decreased with increasing deformation temperature and decreasing strain rate, while flow stress increased with increasing Cu content for all deformation conditions studied due to the solute drag effect. Based on the experimental data, an artificial neural network (ANN) model was developed to study the relationship between chemical composition, deformation variables and high-temperature flow behavior. A three-layer feed-forward back-propagation artificial neural network with 20 neurons in a hidden layer was established in this study. The input parameters were Cu content, temperature, strain rate and strain, while the flow stress was the output. The performance of the proposed model was evaluated using the K-fold cross-validation method. The results showed excellent generalization capability of the developed model. Sensitivity analysis indicated that the strain rate is the most important parameter, while the Cu content exhibited a modest but significant influence on the flow stress.

Modeling the Effects of Cu Content and Deformation Variables on the High-Temperature Flow Behavior of Dilute Al-Fe-Si Alloys Using an Artificial Neural Network

PubMed Central

Shakiba, Mohammad; Parson, Nick; Chen, X.-Grant

2016-01-01

The hot deformation behavior of Al-0.12Fe-0.1Si alloys with varied amounts of Cu (0.002–0.31 wt %) was investigated by uniaxial compression tests conducted at different temperatures (400 °C–550 °C) and strain rates (0.01–10 s−1). The results demonstrated that flow stress decreased with increasing deformation temperature and decreasing strain rate, while flow stress increased with increasing Cu content for all deformation conditions studied due to the solute drag effect. Based on the experimental data, an artificial neural network (ANN) model was developed to study the relationship between chemical composition, deformation variables and high-temperature flow behavior. A three-layer feed-forward back-propagation artificial neural network with 20 neurons in a hidden layer was established in this study. The input parameters were Cu content, temperature, strain rate and strain, while the flow stress was the output. The performance of the proposed model was evaluated using the K-fold cross-validation method. The results showed excellent generalization capability of the developed model. Sensitivity analysis indicated that the strain rate is the most important parameter, while the Cu content exhibited a modest but significant influence on the flow stress. PMID:28773658

A Novel Extreme Learning Machine Classification Model for e-Nose Application Based on the Multiple Kernel Approach

PubMed Central

Jian, Yulin; Huang, Daoyu; Yan, Jia; Lu, Kun; Huang, Ying; Wen, Tailai; Zeng, Tanyue; Zhong, Shijie; Xie, Qilong

2017-01-01

A novel classification model, named the quantum-behaved particle swarm optimization (QPSO)-based weighted multiple kernel extreme learning machine (QWMK-ELM), is proposed in this paper. Experimental validation is carried out with two different electronic nose (e-nose) datasets. Being different from the existing multiple kernel extreme learning machine (MK-ELM) algorithms, the combination coefficients of base kernels are regarded as external parameters of single-hidden layer feedforward neural networks (SLFNs). The combination coefficients of base kernels, the model parameters of each base kernel, and the regularization parameter are optimized by QPSO simultaneously before implementing the kernel extreme learning machine (KELM) with the composite kernel function. Four types of common single kernel functions (Gaussian kernel, polynomial kernel, sigmoid kernel, and wavelet kernel) are utilized to constitute different composite kernel functions. Moreover, the method is also compared with other existing classification methods: extreme learning machine (ELM), kernel extreme learning machine (KELM), k-nearest neighbors (KNN), support vector machine (SVM), multi-layer perceptron (MLP), radical basis function neural network (RBFNN), and probabilistic neural network (PNN). The results have demonstrated that the proposed QWMK-ELM outperforms the aforementioned methods, not only in precision, but also in efficiency for gas classification. PMID:28629202

Optimization of matrix tablets controlled drug release using Elman dynamic neural networks and decision trees.

PubMed

Petrović, Jelena; Ibrić, Svetlana; Betz, Gabriele; Đurić, Zorica

2012-05-30

The main objective of the study was to develop artificial intelligence methods for optimization of drug release from matrix tablets regardless of the matrix type. Static and dynamic artificial neural networks of the same topology were developed to model dissolution profiles of different matrix tablets types (hydrophilic/lipid) using formulation composition, compression force used for tableting and tablets porosity and tensile strength as input data. Potential application of decision trees in discovering knowledge from experimental data was also investigated. Polyethylene oxide polymer and glyceryl palmitostearate were used as matrix forming materials for hydrophilic and lipid matrix tablets, respectively whereas selected model drugs were diclofenac sodium and caffeine. Matrix tablets were prepared by direct compression method and tested for in vitro dissolution profiles. Optimization of static and dynamic neural networks used for modeling of drug release was performed using Monte Carlo simulations or genetic algorithms optimizer. Decision trees were constructed following discretization of data. Calculated difference (f(1)) and similarity (f(2)) factors for predicted and experimentally obtained dissolution profiles of test matrix tablets formulations indicate that Elman dynamic neural networks as well as decision trees are capable of accurate predictions of both hydrophilic and lipid matrix tablets dissolution profiles. Elman neural networks were compared to most frequently used static network, Multi-layered perceptron, and superiority of Elman networks have been demonstrated. Developed methods allow simple, yet very precise way of drug release predictions for both hydrophilic and lipid matrix tablets having controlled drug release. Copyright © 2012 Elsevier B.V. All rights reserved.

Joint multiple fully connected convolutional neural network with extreme learning machine for hepatocellular carcinoma nuclei grading.

PubMed

Li, Siqi; Jiang, Huiyan; Pang, Wenbo

2017-05-01

Accurate cell grading of cancerous tissue pathological image is of great importance in medical diagnosis and treatment. This paper proposes a joint multiple fully connected convolutional neural network with extreme learning machine (MFC-CNN-ELM) architecture for hepatocellular carcinoma (HCC) nuclei grading. First, in preprocessing stage, each grayscale image patch with the fixed size is obtained using center-proliferation segmentation (CPS) method and the corresponding labels are marked under the guidance of three pathologists. Next, a multiple fully connected convolutional neural network (MFC-CNN) is designed to extract the multi-form feature vectors of each input image automatically, which considers multi-scale contextual information of deep layer maps sufficiently. After that, a convolutional neural network extreme learning machine (CNN-ELM) model is proposed to grade HCC nuclei. Finally, a back propagation (BP) algorithm, which contains a new up-sample method, is utilized to train MFC-CNN-ELM architecture. The experiment comparison results demonstrate that our proposed MFC-CNN-ELM has superior performance compared with related works for HCC nuclei grading. Meanwhile, external validation using ICPR 2014 HEp-2 cell dataset shows the good generalization of our MFC-CNN-ELM architecture. Copyright © 2017 Elsevier Ltd. All rights reserved.

Hopfield neural network and optical fiber sensor as intelligent heart rate monitor

NASA Astrophysics Data System (ADS)

Mutter, Kussay Nugamesh

2018-01-01

This paper presents a design and fabrication of an intelligent fiber-optic sensor used for examining and monitoring heart rate activity. It is found in the literature that the use of fiber sensors as heart rate sensor is widely studied. However, the use of smart sensors based on Hopfield neural networks is very low. In this work, the sensor is a three fibers without cladding of about 1 cm, fed by laser light of 1550 nm of wavelength. The sensing portions are mounted with a micro sensitive diaphragm to transfer the pulse pressure on the left radial wrist. The influenced light intensity will be detected by a three photodetectors as inputs into the Hopfield neural network algorithm. The latter is a singlelayer auto-associative memory structure with a same input and output layers. The prior training weights are stored in the net memory for the standard recorded normal heart rate signals. The sensors' heads work on the reflection intensity basis. The novelty here is that the sensor uses a pulse pressure and Hopfield neural network in an integrity approach. The results showed a significant output measurements of heart rate and counting with a plausible error rate.

Detection of inter-turn short-circuit at start-up of induction machine based on torque analysis

NASA Astrophysics Data System (ADS)

Pietrowski, Wojciech; Górny, Konrad

2017-12-01

Recently, interest in new diagnostics methods in a field of induction machines was observed. Research presented in the paper shows the diagnostics of induction machine based on torque pulsation, under inter-turn short-circuit, during start-up of a machine. In the paper three numerical techniques were used: finite element analysis, signal analysis and artificial neural networks (ANN). The elaborated numerical model of faulty machine consists of field, circuit and motion equations. Voltage excited supply allowed to determine the torque waveform during start-up. The inter-turn short-circuit was treated as a galvanic connection between two points of the stator winding. The waveforms were calculated for different amounts of shorted-turns from 0 to 55. Due to the non-stationary waveforms a wavelet packet decomposition was used to perform an analysis of the torque. The obtained results of analysis were used as input vector for ANN. The response of the neural network was the number of shorted-turns in the stator winding. Special attention was paid to compare response of general regression neural network (GRNN) and multi-layer perceptron neural network (MLP). Based on the results of the research, the efficiency of the developed algorithm can be inferred.

Analysis of Mars Express Ionogram Data via a Multilayer Artificial Neural Network

NASA Astrophysics Data System (ADS)

Wilkinson, Collin; Potter, Arron; Palmer, Greg; Duru, Firdevs

2017-01-01

Mars Advanced Radar for Subsurface and Ionospheric Sounding (MARSIS), which is a low frequency radar on the Mars Express (MEX) Spacecraft, can provide electron plasma densities of the ionosphere local at the spacecraft in addition to densities obtained with remote sounding. The local electron densities are obtained, with a standard error of about 2%, by measuring the electron plasma frequencies with an electronic ruler on ionograms, which are plots of echo intensity as a function of time and frequency. This is done by using a tool created at the University of Iowa (Duru et al., 2008). This approach is time consuming due to the rapid accumulation of ionogram data. In 2013, results from an algorithm-based analysis of ionograms were reported by Andrews et al., but this method did not improve the human error. In the interest of fast, accurate data interpretation, a neural network (NN) has been created based on the Fast Artificial Neural Network C libraries. This NN consists of artificial neurons, with 4 layers of 12960, 10000, 1000 and 1 neuron(s) each, consecutively. This network was trained using 40 iterations of 1000 orbits. The algorithm-based method of Andrews et al. had a standard error of 40%, while the neural network has achieved error on the order of 20%.

A Viola-Jones based hybrid face detection framework

NASA Astrophysics Data System (ADS)

Murphy, Thomas M.; Broussard, Randy; Schultz, Robert; Rakvic, Ryan; Ngo, Hau

2013-12-01

Improvements in face detection performance would benefit many applications. The OpenCV library implements a standard solution, the Viola-Jones detector, with a statistically boosted rejection cascade of binary classifiers. Empirical evidence has shown that Viola-Jones underdetects in some instances. This research shows that a truncated cascade augmented by a neural network could recover these undetected faces. A hybrid framework is constructed, with a truncated Viola-Jones cascade followed by an artificial neural network, used to refine the face decision. Optimally, a truncation stage that captured all faces and allowed the neural network to remove the false alarms is selected. A feedforward backpropagation network with one hidden layer is trained to discriminate faces based upon the thresholding (detection) values of intermediate stages of the full rejection cascade. A clustering algorithm is used as a precursor to the neural network, to group significant overlappings. Evaluated on the CMU/VASC Image Database, comparison with an unmodified OpenCV approach shows: (1) a 37% increase in detection rates if constrained by the requirement of no increase in false alarms, (2) a 48% increase in detection rates if some additional false alarms are tolerated, and (3) an 82% reduction in false alarms with no reduction in detection rates. These results demonstrate improved face detection and could address the need for such improvement in various applications.

Correcting wave predictions with artificial neural networks

NASA Astrophysics Data System (ADS)

Makarynskyy, O.; Makarynska, D.

2003-04-01

The predictions of wind waves with different lead times are necessary in a large scope of coastal and open ocean activities. Numerical wave models, which usually provide this information, are based on deterministic equations that do not entirely account for the complexity and uncertainty of the wave generation and dissipation processes. An attempt to improve wave parameters short-term forecasts based on artificial neural networks is reported. In recent years, artificial neural networks have been used in a number of coastal engineering applications due to their ability to approximate the nonlinear mathematical behavior without a priori knowledge of interrelations among the elements within a system. The common multilayer feed-forward networks, with a nonlinear transfer functions in the hidden layers, were developed and employed to forecast the wave characteristics over one hour intervals starting from one up to 24 hours, and to correct these predictions. Three non-overlapping data sets of wave characteristics, both from a buoy, moored roughly 60 miles west of the Aran Islands, west coast of Ireland, were used to train and validate the neural nets involved. The networks were trained with error back propagation algorithm. Time series plots and scatterplots of the wave characteristics as well as tables with statistics show an improvement of the results achieved due to the correction procedure employed.

Neural control of magnetic suspension systems

NASA Technical Reports Server (NTRS)

Gray, W. Steven

1993-01-01

The purpose of this research program is to design, build and test (in cooperation with NASA personnel from the NASA Langley Research Center) neural controllers for two different small air-gap magnetic suspension systems. The general objective of the program is to study neural network architectures for the purpose of control in an experimental setting and to demonstrate the feasibility of the concept. The specific objectives of the research program are: (1) to demonstrate through simulation and experimentation the feasibility of using neural controllers to stabilize a nonlinear magnetic suspension system; (2) to investigate through simulation and experimentation the performance of neural controllers designs under various types of parametric and nonparametric uncertainty; (3) to investigate through simulation and experimentation various types of neural architectures for real-time control with respect to performance and complexity; and (4) to benchmark in an experimental setting the performance of neural controllers against other types of existing linear and nonlinear compensator designs. To date, the first one-dimensional, small air-gap magnetic suspension system has been built, tested and delivered to the NASA Langley Research Center. The device is currently being stabilized with a digital linear phase-lead controller. The neural controller hardware is under construction. Two different neural network paradigms are under consideration, one based on hidden layer feedforward networks trained via back propagation and one based on using Gaussian radial basis functions trained by analytical methods related to stability conditions. Some advanced nonlinear control algorithms using feedback linearization and sliding mode control are in simulation studies.

Fast learning method for convolutional neural networks using extreme learning machine and its application to lane detection.

PubMed

Kim, Jihun; Kim, Jonghong; Jang, Gil-Jin; Lee, Minho

2017-03-01

Deep learning has received significant attention recently as a promising solution to many problems in the area of artificial intelligence. Among several deep learning architectures, convolutional neural networks (CNNs) demonstrate superior performance when compared to other machine learning methods in the applications of object detection and recognition. We use a CNN for image enhancement and the detection of driving lanes on motorways. In general, the process of lane detection consists of edge extraction and line detection. A CNN can be used to enhance the input images before lane detection by excluding noise and obstacles that are irrelevant to the edge detection result. However, training conventional CNNs requires considerable computation and a big dataset. Therefore, we suggest a new learning algorithm for CNNs using an extreme learning machine (ELM). The ELM is a fast learning method used to calculate network weights between output and hidden layers in a single iteration and thus, can dramatically reduce learning time while producing accurate results with minimal training data. A conventional ELM can be applied to networks with a single hidden layer; as such, we propose a stacked ELM architecture in the CNN framework. Further, we modify the backpropagation algorithm to find the targets of hidden layers and effectively learn network weights while maintaining performance. Experimental results confirm that the proposed method is effective in reducing learning time and improving performance. Copyright © 2016 Elsevier Ltd. All rights reserved.

The use of wavelet packet transform and artificial neural networks in analysis and classification of dysphonic voices.

PubMed

Crovato, César David Paredes; Schuck, Adalberto

2007-10-01

This paper presents a dysphonic voice classification system using the wavelet packet transform and the best basis algorithm (BBA) as dimensionality reductor and 06 artificial neural networks (ANN) acting as specialist systems. Each ANN was a 03-layer multilayer perceptron with 64 input nodes, 01 output node and in the intermediary layer the number of neurons depends on the related training pathology group. The dysphonic voice database was separated in five pathology groups and one healthy control group. Each ANN was trained and associated with one of the 06 groups, and fed by the best base tree (BBT) nodes' entropy values, using the multiple cross validation (MCV) method and the leave-one-out (LOO) variation technique and success rates obtained were 87.5%, 95.31%, 87.5%, 100%, 96.87% and 89.06% for the groups 01 to 06, respectively.

[Application of an artificial neural network in the design of sustained-release dosage forms].

PubMed

Wei, X H; Wu, J J; Liang, W Q

2001-09-01

To use the artificial neural network (ANN) in Matlab 5.1 tool-boxes to predict the formulations of sustained-release tablets. The solubilities of nine drugs and various ratios of HPMC: Dextrin for 63 tablet formulations were used as the ANN model input, and in vitro accumulation released at 6 sampling times were used as output. The ANN model was constructed by selecting the optimal number of iterations (25) and model structure in which there are one hidden layer and five hidden layer nodes. The optimized ANN model was used for prediction of formulation based on desired target in vitro dissolution-time profiles. ANN predicted profiles based on ANN predicted formulations were closely similar to the target profiles. The ANN could be used for predicting the dissolution profiles of sustained release dosage form and for the design of optimal formulation.

An Efficient Implementation of Deep Convolutional Neural Networks for MRI Segmentation.

PubMed

Hoseini, Farnaz; Shahbahrami, Asadollah; Bayat, Peyman

2018-02-27

Image segmentation is one of the most common steps in digital image processing, classifying a digital image into different segments. The main goal of this paper is to segment brain tumors in magnetic resonance images (MRI) using deep learning. Tumors having different shapes, sizes, brightness and textures can appear anywhere in the brain. These complexities are the reasons to choose a high-capacity Deep Convolutional Neural Network (DCNN) containing more than one layer. The proposed DCNN contains two parts: architecture and learning algorithms. The architecture and the learning algorithms are used to design a network model and to optimize parameters for the network training phase, respectively. The architecture contains five convolutional layers, all using 3 × 3 kernels, and one fully connected layer. Due to the advantage of using small kernels with fold, it allows making the effect of larger kernels with smaller number of parameters and fewer computations. Using the Dice Similarity Coefficient metric, we report accuracy results on the BRATS 2016, brain tumor segmentation challenge dataset, for the complete, core, and enhancing regions as 0.90, 0.85, and 0.84 respectively. The learning algorithm includes the task-level parallelism. All the pixels of an MR image are classified using a patch-based approach for segmentation. We attain a good performance and the experimental results show that the proposed DCNN increases the segmentation accuracy compared to previous techniques.

A design philosophy for multi-layer neural networks with applications to robot control

NASA Technical Reports Server (NTRS)

Vadiee, Nader; Jamshidi, MO

1989-01-01

A system is proposed which receives input information from many sensors that may have diverse scaling, dimension, and data representations. The proposed system tolerates sensory information with faults. The proposed self-adaptive processing technique has great promise in integrating the techniques of artificial intelligence and neural networks in an attempt to build a more intelligent computing environment. The proposed architecture can provide a detailed decision tree based on the input information, information stored in a long-term memory, and the adapted rule-based knowledge. A mathematical model for analysis will be obtained to validate the cited hypotheses. An extensive software program will be developed to simulate a typical example of pattern recognition problem. It is shown that the proposed model displays attention, expectation, spatio-temporal, and predictory behavior which are specific to the human brain. The anticipated results of this research project are: (1) creation of a new dynamic neural network structure, and (2) applications to and comparison with conventional multi-layer neural network structures. The anticipated benefits from this research are vast. The model can be used in a neuro-computer architecture as a building block which can perform complicated, nonlinear, time-varying mapping from a multitude of input excitory classes to an output or decision environment. It can be used for coordinating different sensory inputs and past experience of a dynamic system and actuating signals. The commercial applications of this project can be the creation of a special-purpose neuro-computer hardware which can be used in spatio-temporal pattern recognitions in such areas as air defense systems, e.g., target tracking, and recognition. Potential robotics-related applications are trajectory planning, inverse dynamics computations, hierarchical control, task-oriented control, and collision avoidance.

Wearable-Sensor-Based Classification Models of Faller Status in Older Adults.

PubMed

Howcroft, Jennifer; Lemaire, Edward D; Kofman, Jonathan

2016-01-01

Wearable sensors have potential for quantitative, gait-based, point-of-care fall risk assessment that can be easily and quickly implemented in clinical-care and older-adult living environments. This investigation generated models for wearable-sensor based fall-risk classification in older adults and identified the optimal sensor type, location, combination, and modelling method; for walking with and without a cognitive load task. A convenience sample of 100 older individuals (75.5 ± 6.7 years; 76 non-fallers, 24 fallers based on 6 month retrospective fall occurrence) walked 7.62 m under single-task and dual-task conditions while wearing pressure-sensing insoles and tri-axial accelerometers at the head, pelvis, and left and right shanks. Participants also completed the Activities-specific Balance Confidence scale, Community Health Activities Model Program for Seniors questionnaire, six minute walk test, and ranked their fear of falling. Fall risk classification models were assessed for all sensor combinations and three model types: multi-layer perceptron neural network, naïve Bayesian, and support vector machine. The best performing model was a multi-layer perceptron neural network with input parameters from pressure-sensing insoles and head, pelvis, and left shank accelerometers (accuracy = 84%, F1 score = 0.600, MCC score = 0.521). Head sensor-based models had the best performance of the single-sensor models for single-task gait assessment. Single-task gait assessment models outperformed models based on dual-task walking or clinical assessment data. Support vector machines and neural networks were the best modelling technique for fall risk classification. Fall risk classification models developed for point-of-care environments should be developed using support vector machines and neural networks, with a multi-sensor single-task gait assessment.

Generalized Predictive and Neural Generalized Predictive Control of Aerospace Systems

NASA Technical Reports Server (NTRS)

Kelkar, Atul G.

2000-01-01

The research work presented in this thesis addresses the problem of robust control of uncertain linear and nonlinear systems using Neural network-based Generalized Predictive Control (NGPC) methodology. A brief overview of predictive control and its comparison with Linear Quadratic (LQ) control is given to emphasize advantages and drawbacks of predictive control methods. It is shown that the Generalized Predictive Control (GPC) methodology overcomes the drawbacks associated with traditional LQ control as well as conventional predictive control methods. It is shown that in spite of the model-based nature of GPC it has good robustness properties being special case of receding horizon control. The conditions for choosing tuning parameters for GPC to ensure closed-loop stability are derived. A neural network-based GPC architecture is proposed for the control of linear and nonlinear uncertain systems. A methodology to account for parametric uncertainty in the system is proposed using on-line training capability of multi-layer neural network. Several simulation examples and results from real-time experiments are given to demonstrate the effectiveness of the proposed methodology.

Three-dimensional micro-electrode array for recording dissociated neuronal cultures.

PubMed

Musick, Katherine; Khatami, David; Wheeler, Bruce C

2009-07-21

This work demonstrates the design, fabrication, packaging, characterization, and functionality of an electrically and fluidically active three-dimensional micro-electrode array (3D MEA) for use with neuronal cell cultures. The successful function of the device implies that this basic concept-construction of a 3D array with a layered approach-can be utilized as the basis for a new family of neural electrode arrays. The 3D MEA prototype consists of a stack of individually patterned thin films that form a cell chamber conducive to maintaining and recording the electrical activity of a long-term three-dimensional network of rat cortical neurons. Silicon electrode layers contain a polymer grid for neural branching, growth, and network formation. Along the walls of these electrode layers lie exposed gold electrodes which permit recording and stimulation of the neuronal electrical activity. Silicone elastomer micro-fluidic layers provide a means for loading dissociated neurons into the structure and serve as the artificial vasculature for nutrient supply and aeration. The fluidic layers also serve as insulation for the micro-electrodes. Cells have been shown to survive in the 3D MEA for up to 28 days, with spontaneous and evoked electrical recordings performed in that time. The micro-fluidic capability was demonstrated by flowing in the drug tetrotodoxin to influence the activity of the culture.

Synthesizing cognition in neuromorphic electronic systems

PubMed Central

Neftci, Emre; Binas, Jonathan; Rutishauser, Ueli; Chicca, Elisabetta; Indiveri, Giacomo; Douglas, Rodney J.

2013-01-01

The quest to implement intelligent processing in electronic neuromorphic systems lacks methods for achieving reliable behavioral dynamics on substrates of inherently imprecise and noisy neurons. Here we report a solution to this problem that involves first mapping an unreliable hardware layer of spiking silicon neurons into an abstract computational layer composed of generic reliable subnetworks of model neurons and then composing the target behavioral dynamics as a “soft state machine” running on these reliable subnets. In the first step, the neural networks of the abstract layer are realized on the hardware substrate by mapping the neuron circuit bias voltages to the model parameters. This mapping is obtained by an automatic method in which the electronic circuit biases are calibrated against the model parameters by a series of population activity measurements. The abstract computational layer is formed by configuring neural networks as generic soft winner-take-all subnetworks that provide reliable processing by virtue of their active gain, signal restoration, and multistability. The necessary states and transitions of the desired high-level behavior are then easily embedded in the computational layer by introducing only sparse connections between some neurons of the various subnets. We demonstrate this synthesis method for a neuromorphic sensory agent that performs real-time context-dependent classification of motion patterns observed by a silicon retina. PMID:23878215

Bounds on the number of hidden neurons in three-layer binary neural networks.

PubMed

Zhang, Zhaozhi; Ma, Xiaomin; Yang, Yixian

2003-09-01

This paper investigates an important problem concerning the complexity of three-layer binary neural networks (BNNs) with one hidden layer. The neuron in the studied BNNs employs a hard limiter activation function with only integer weights and an integer threshold. The studies are focused on implementations of arbitrary Boolean functions which map from [0, 1]n into [0, 1]. A deterministic algorithm called set covering algorithm (SCA) is proposed for the construction of a three-layer BNN to implement an arbitrary Boolean function. The SCA is based on a unit sphere covering (USC) of the Hamming space (HS) which is chosen in advance. It is proved that for the implementation of an arbitrary Boolean function of n-variables (n > or = 3) by using SCA, [3L/2] hidden neurons are necessary and sufficient, where L is the number of unit spheres contained in the chosen USC of the n-dimensional HS. It is shown that by using SCA, the number of hidden neurons required is much less than that by using a two-parallel hyperplane method. In order to indicate the potential ability of three-layer BNNs, a lower bound on the required number of hidden neurons which is derived by using the method of estimating the Vapnik-Chervonenkis (VC) dimension is also given.

[Terahertz Spectroscopic Identification with Deep Belief Network].

PubMed

Ma, Shuai; Shen, Tao; Wang, Rui-qi; Lai, Hua; Yu, Zheng-tao

2015-12-01

Feature extraction and classification are the key issues of terahertz spectroscopy identification. Because many materials have no apparent absorption peaks in the terahertz band, it is difficult to extract theirs terahertz spectroscopy feature and identify. To this end, a novel of identify terahertz spectroscopy approach with Deep Belief Network (DBN) was studied in this paper, which combines the advantages of DBN and K-Nearest Neighbors (KNN) classifier. Firstly, cubic spline interpolation and S-G filter were used to normalize the eight kinds of substances (ATP, Acetylcholine Bromide, Bifenthrin, Buprofezin, Carbazole, Bleomycin, Buckminster and Cylotriphosphazene) terahertz transmission spectra in the range of 0.9-6 THz. Secondly, the DBN model was built by two restricted Boltzmann machine (RBM) and then trained layer by layer using unsupervised approach. Instead of using handmade features, the DBN was employed to learn suitable features automatically with raw input data. Finally, a KNN classifier was applied to identify the terahertz spectrum. Experimental results show that using the feature learned by DBN can identify the terahertz spectrum of different substances with the recognition rate of over 90%, which demonstrates that the proposed method can automatically extract the effective features of terahertz spectrum. Furthermore, this KNN classifier was compared with others (BP neural network, SOM neural network and RBF neural network). Comparisons showed that the recognition rate of KNN classifier is better than the other three classifiers. Using the approach that automatic extract terahertz spectrum features by DBN can greatly reduce the workload of feature extraction. This proposed method shows a promising future in the application of identifying the mass terahertz spectroscopy.

Brain tumor segmentation with Deep Neural Networks.

PubMed

Havaei, Mohammad; Davy, Axel; Warde-Farley, David; Biard, Antoine; Courville, Aaron; Bengio, Yoshua; Pal, Chris; Jodoin, Pierre-Marc; Larochelle, Hugo

2017-01-01

In this paper, we present a fully automatic brain tumor segmentation method based on Deep Neural Networks (DNNs). The proposed networks are tailored to glioblastomas (both low and high grade) pictured in MR images. By their very nature, these tumors can appear anywhere in the brain and have almost any kind of shape, size, and contrast. These reasons motivate our exploration of a machine learning solution that exploits a flexible, high capacity DNN while being extremely efficient. Here, we give a description of different model choices that we've found to be necessary for obtaining competitive performance. We explore in particular different architectures based on Convolutional Neural Networks (CNN), i.e. DNNs specifically adapted to image data. We present a novel CNN architecture which differs from those traditionally used in computer vision. Our CNN exploits both local features as well as more global contextual features simultaneously. Also, different from most traditional uses of CNNs, our networks use a final layer that is a convolutional implementation of a fully connected layer which allows a 40 fold speed up. We also describe a 2-phase training procedure that allows us to tackle difficulties related to the imbalance of tumor labels. Finally, we explore a cascade architecture in which the output of a basic CNN is treated as an additional source of information for a subsequent CNN. Results reported on the 2013 BRATS test data-set reveal that our architecture improves over the currently published state-of-the-art while being over 30 times faster. Copyright © 2016 Elsevier B.V. All rights reserved.

Neural Network Technique for Continous Transition from Ocean to Coastal Retrackers

NASA Astrophysics Data System (ADS)

Hazrina Idris, Nurul; Deng, Xiaoli; Hawani Idris, Nurul

2017-04-01

This paper presents the development of neural network for continuous transition of altimeter sea surface heights when switching from ocean to coastal waveform retrackers. In attempting to produce precise coastal sea level anomaly (SLA) via retracking waveforms, issue arose when employing multiple retracking algorithms (i.e. MLE-4, sub-waveform and threshold). The existence of relative offset between those retrackers creates 'jump' in the retracked SLA profiles. In this study, the offset between retrackers is minimized using multi-layer feed forward neural network technique. The technique reduces the offset values by modelling the complicated functions of those retracked SLAs. The technique is tested over the region of the Great Barrier Reef (GBR), Australia. The validation with Townsville and Bundaberg tide gauges shows that the threshold retracker achieves temporal correlations (r) of 0.84 and 0.75, respectively, and root mean square (RMS) error is 16 cm for both stations, indicating that the retracker produces more accurate SLAs than those of two retrackers. Meanwhile, values of r (RMS error) for MLE-4 is only 0.79 (18 cm) and 0.71 (16 cm), respectively, and for sub-waveform is 0.82 (16 cm) and 0.67 (16 cm), respectively. Therefore, with the neural network, retracked SLAs from MLE-4 and sub-waveform are aligned to those of the threshold retracker. The performance of neural network is compared with the normal procedure of offset removal, which is based on the mean of SLA differences (mean method). The performance is assessed by computing the standard deviation of difference (STD) between the SLAs above a referenced ellipsoid and the geoidal height, and the improvement of percentage (IMP). The results indicate that the neural network provides improvement in SLA precision in all 12 cases, while the mean method provides improvement in 10 out of 12 cases and deterioration is seen in two cases. In terms of STD and IMP, neural network reduces the offset better than those of the mean method. The IMPs with neural network reaches up to 67% for Jason-1 and 73% for Jason-2, meanwhile with mean method the IMPs only reaches up to 28% and 46%, respectively. In conclusion, the neural network technique is efficient to reduce the offset among retrackers by handling the linear and nonlinear relationship between retrackers, thus providing seamless transition from the open ocean to the coast, and vice versa. Studies in currently on-going are to consider other geophysical parameters, such as significant wave height that might be related to the variation of the offset, in the neural network.

Event Recognition Based on Deep Learning in Chinese Texts

PubMed Central

Zhang, Yajun; Liu, Zongtian; Zhou, Wen

2016-01-01

Event recognition is the most fundamental and critical task in event-based natural language processing systems. Existing event recognition methods based on rules and shallow neural networks have certain limitations. For example, extracting features using methods based on rules is difficult; methods based on shallow neural networks converge too quickly to a local minimum, resulting in low recognition precision. To address these problems, we propose the Chinese emergency event recognition model based on deep learning (CEERM). Firstly, we use a word segmentation system to segment sentences. According to event elements labeled in the CEC 2.0 corpus, we classify words into five categories: trigger words, participants, objects, time and location. Each word is vectorized according to the following six feature layers: part of speech, dependency grammar, length, location, distance between trigger word and core word and trigger word frequency. We obtain deep semantic features of words by training a feature vector set using a deep belief network (DBN), then analyze those features in order to identify trigger words by means of a back propagation neural network. Extensive testing shows that the CEERM achieves excellent recognition performance, with a maximum F-measure value of 85.17%. Moreover, we propose the dynamic-supervised DBN, which adds supervised fine-tuning to a restricted Boltzmann machine layer by monitoring its training performance. Test analysis reveals that the new DBN improves recognition performance and effectively controls the training time. Although the F-measure increases to 88.11%, the training time increases by only 25.35%. PMID:27501231

Event Recognition Based on Deep Learning in Chinese Texts.

PubMed

Zhang, Yajun; Liu, Zongtian; Zhou, Wen

2016-01-01

Event recognition is the most fundamental and critical task in event-based natural language processing systems. Existing event recognition methods based on rules and shallow neural networks have certain limitations. For example, extracting features using methods based on rules is difficult; methods based on shallow neural networks converge too quickly to a local minimum, resulting in low recognition precision. To address these problems, we propose the Chinese emergency event recognition model based on deep learning (CEERM). Firstly, we use a word segmentation system to segment sentences. According to event elements labeled in the CEC 2.0 corpus, we classify words into five categories: trigger words, participants, objects, time and location. Each word is vectorized according to the following six feature layers: part of speech, dependency grammar, length, location, distance between trigger word and core word and trigger word frequency. We obtain deep semantic features of words by training a feature vector set using a deep belief network (DBN), then analyze those features in order to identify trigger words by means of a back propagation neural network. Extensive testing shows that the CEERM achieves excellent recognition performance, with a maximum F-measure value of 85.17%. Moreover, we propose the dynamic-supervised DBN, which adds supervised fine-tuning to a restricted Boltzmann machine layer by monitoring its training performance. Test analysis reveals that the new DBN improves recognition performance and effectively controls the training time. Although the F-measure increases to 88.11%, the training time increases by only 25.35%.