Pub Date : 1999-07-10DOI: 10.1109/IJCNN.1999.831507
A. Burkitt
The observed variability in the spike rate of cortical neurons has been hypothesized to result from a balance in the excitatory and inhibitory synaptic inputs that the neurons receive. The coefficient of variation of the spike rate is calculated in the limit of a large number of inputs using the integrated-input technique, which is extended here to include the effect of reversal potentials. The output spike rate is found to increase monotonically over two orders of magnitude, thereby solving the dynamic range (or gain control) problem. The coefficient of variation is approximately 1.0 for low input rates and increases to around 1.6 at high input rates, well within the range observed in the response of cortical neurons.
{"title":"Analysis of neural response for excitation-inhibition balanced networks with reversal potentials for large numbers of inputs","authors":"A. Burkitt","doi":"10.1109/IJCNN.1999.831507","DOIUrl":"https://doi.org/10.1109/IJCNN.1999.831507","url":null,"abstract":"The observed variability in the spike rate of cortical neurons has been hypothesized to result from a balance in the excitatory and inhibitory synaptic inputs that the neurons receive. The coefficient of variation of the spike rate is calculated in the limit of a large number of inputs using the integrated-input technique, which is extended here to include the effect of reversal potentials. The output spike rate is found to increase monotonically over two orders of magnitude, thereby solving the dynamic range (or gain control) problem. The coefficient of variation is approximately 1.0 for low input rates and increases to around 1.6 at high input rates, well within the range observed in the response of cortical neurons.","PeriodicalId":157719,"journal":{"name":"IJCNN'99. International Joint Conference on Neural Networks. Proceedings (Cat. No.99CH36339)","volume":"198 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134105756","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1999-07-10DOI: 10.1109/IJCNN.1999.831571
Shi Zhong, V. Cherkassky
Multilayer perceptron (MLP) network has been successfully applied to many practical problems because of its nonlinear mapping ability. However, there are many factors, which may affect the generalization ability of MLP networks, such as the number of hidden units, the initial values of weights and the stopping rules. These factors, if improperly chosen, may result in poor generalization ability of MLP networks. It is important to identify, these factors and their interaction in order to control effectively the generalization ability of MLP network. In this paper, we have empirically identified the factors that affect the generalization ability of MLP network, and compared their relative effect on the generalization performance.
{"title":"Factors controlling generalization ability of MLP networks","authors":"Shi Zhong, V. Cherkassky","doi":"10.1109/IJCNN.1999.831571","DOIUrl":"https://doi.org/10.1109/IJCNN.1999.831571","url":null,"abstract":"Multilayer perceptron (MLP) network has been successfully applied to many practical problems because of its nonlinear mapping ability. However, there are many factors, which may affect the generalization ability of MLP networks, such as the number of hidden units, the initial values of weights and the stopping rules. These factors, if improperly chosen, may result in poor generalization ability of MLP networks. It is important to identify, these factors and their interaction in order to control effectively the generalization ability of MLP network. In this paper, we have empirically identified the factors that affect the generalization ability of MLP network, and compared their relative effect on the generalization performance.","PeriodicalId":157719,"journal":{"name":"IJCNN'99. International Joint Conference on Neural Networks. Proceedings (Cat. No.99CH36339)","volume":"76 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134374661","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1999-07-10DOI: 10.1109/IJCNN.1999.836002
Lin Zhong, Yuanyuan Shi, Runsheng Liu
A dynamic neural network architecture based on the time-delay neural network and the convolutional neural network is originated. The dynamic network achieves much better performance than those of MLP and TDNN when dealing with syllable recognition. Such performance is also comparable to that of the more popular HMM method.
{"title":"A dynamic neural network for syllable recognition","authors":"Lin Zhong, Yuanyuan Shi, Runsheng Liu","doi":"10.1109/IJCNN.1999.836002","DOIUrl":"https://doi.org/10.1109/IJCNN.1999.836002","url":null,"abstract":"A dynamic neural network architecture based on the time-delay neural network and the convolutional neural network is originated. The dynamic network achieves much better performance than those of MLP and TDNN when dealing with syllable recognition. Such performance is also comparable to that of the more popular HMM method.","PeriodicalId":157719,"journal":{"name":"IJCNN'99. International Joint Conference on Neural Networks. Proceedings (Cat. No.99CH36339)","volume":"519 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131655167","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1999-07-10DOI: 10.1109/IJCNN.1999.831577
S. Belbas
We examine certain aspects of neural networks based on controlled 2D systems representable as ordinary differential equations: the use of the asymptotic behaviour of 2D dynamical systems for optimal design, and the use of optimal control of 2D controlled systems for devising learning strategies.
{"title":"Neural networks in 2-D continuous time","authors":"S. Belbas","doi":"10.1109/IJCNN.1999.831577","DOIUrl":"https://doi.org/10.1109/IJCNN.1999.831577","url":null,"abstract":"We examine certain aspects of neural networks based on controlled 2D systems representable as ordinary differential equations: the use of the asymptotic behaviour of 2D dynamical systems for optimal design, and the use of optimal control of 2D controlled systems for devising learning strategies.","PeriodicalId":157719,"journal":{"name":"IJCNN'99. International Joint Conference on Neural Networks. Proceedings (Cat. No.99CH36339)","volume":"25 13","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132835981","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1999-07-10DOI: 10.1109/IJCNN.1999.832678
P. Hannah, R. Stonier, C. Cole
We demonstrate the self-organising continuous map (SOCM), a novel use for the self-organising map/learning vector quantisation network that widens the scope of the SOM architecture. We use the SOM/LVQ network as a distribution service, apportioning an equal quantity of work to a number of intelligent nodes. Advantages include improved accuracy, effective and balanced multi-processing for small cluster systems, and potentially large reductions in training and recall times. The example problem chosen uses neural networks to model force dynamics of a coal train. The SOCM configuration used consists of a SOM network where each node is a backpropagation (BP) network. We show that the collection of as few as two BP networks gives at least a 30% reduction in approximation error when compared to the original BP network. We discuss how the SOCM approach could be used in other areas of artificial intelligence, including evolutionary systems, parallel processing, error balancing, hybrid networks, and online training.
{"title":"Approximating rail locomotive dynamics using the SOCM network","authors":"P. Hannah, R. Stonier, C. Cole","doi":"10.1109/IJCNN.1999.832678","DOIUrl":"https://doi.org/10.1109/IJCNN.1999.832678","url":null,"abstract":"We demonstrate the self-organising continuous map (SOCM), a novel use for the self-organising map/learning vector quantisation network that widens the scope of the SOM architecture. We use the SOM/LVQ network as a distribution service, apportioning an equal quantity of work to a number of intelligent nodes. Advantages include improved accuracy, effective and balanced multi-processing for small cluster systems, and potentially large reductions in training and recall times. The example problem chosen uses neural networks to model force dynamics of a coal train. The SOCM configuration used consists of a SOM network where each node is a backpropagation (BP) network. We show that the collection of as few as two BP networks gives at least a 30% reduction in approximation error when compared to the original BP network. We discuss how the SOCM approach could be used in other areas of artificial intelligence, including evolutionary systems, parallel processing, error balancing, hybrid networks, and online training.","PeriodicalId":157719,"journal":{"name":"IJCNN'99. International Joint Conference on Neural Networks. Proceedings (Cat. No.99CH36339)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129374428","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1999-07-10DOI: 10.1109/IJCNN.1999.836174
H. Abdelbaki, E. Gelenbe, T. Koçak
Remedial mine detection and the detection of unexploded ordnance (UXO) have become very important for humanitarian reasons. This paper addresses mine detection using commonly used electromagnetic induction sensors. We propose and evaluate two neural network approaches to mine detection which provide a robust nonparametric technique, based on training the networks using data from a previously calibrated portion of the minefield, or from a similar minefield. In the first approach, we combine a novel statistic, the S-statistic (which is a real valued variable related to the relative energy difference measured around a point in the minefield) with the /spl delta/-technique in a random neural network (RNN) design. In the second approach, a RNN is trained using a 3/spl times/3 block measurement window, and then applied as a postprocessor for the /spl delta/-technique. This RNN has an unconventional feedforward structure which realizes a matched filter to discriminate between nonmine patterns and mines. Experimental results for both approaches show that the RNN reduces false alarms substantially over the /spl delta/-technique and the energy detector.
{"title":"Matched neural filters for EMI based mine detection","authors":"H. Abdelbaki, E. Gelenbe, T. Koçak","doi":"10.1109/IJCNN.1999.836174","DOIUrl":"https://doi.org/10.1109/IJCNN.1999.836174","url":null,"abstract":"Remedial mine detection and the detection of unexploded ordnance (UXO) have become very important for humanitarian reasons. This paper addresses mine detection using commonly used electromagnetic induction sensors. We propose and evaluate two neural network approaches to mine detection which provide a robust nonparametric technique, based on training the networks using data from a previously calibrated portion of the minefield, or from a similar minefield. In the first approach, we combine a novel statistic, the S-statistic (which is a real valued variable related to the relative energy difference measured around a point in the minefield) with the /spl delta/-technique in a random neural network (RNN) design. In the second approach, a RNN is trained using a 3/spl times/3 block measurement window, and then applied as a postprocessor for the /spl delta/-technique. This RNN has an unconventional feedforward structure which realizes a matched filter to discriminate between nonmine patterns and mines. Experimental results for both approaches show that the RNN reduces false alarms substantially over the /spl delta/-technique and the energy detector.","PeriodicalId":157719,"journal":{"name":"IJCNN'99. International Joint Conference on Neural Networks. Proceedings (Cat. No.99CH36339)","volume":"34 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130737097","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1999-07-10DOI: 10.1109/IJCNN.1999.830835
B. Zheng, Yuan-Hsiang Chang, Xiao-Hui Wang, W. Good
Artificial neural networks (ANN) have been widely used in computer-assisted diagnosis (CAD) schemes as a classification tool to identify abnormalities in digitized mammograms. Because of certain limitations of ANNs, some investigators argue that Bayesian belief network (BBN) may exhibit higher performance. In this study we compared the performance of an ANN and a BBN used in the same CAD scheme. The common databases and the same genetic algorithm (GA) were used to optimize both networks. The experimental results demonstrated that using GA optimization, the performance of the two networks converged to the same level in detecting masses from digitized mammograms. Therefore, in this study we concluded that improving the performance of CAD schemes might be more dependent on optimization of feature selection and diversity of training database than on any particular machine classification paradigm.
{"title":"Comparison of artificial neural network and Bayesian belief network in a computer-assisted diagnosis scheme for mammography","authors":"B. Zheng, Yuan-Hsiang Chang, Xiao-Hui Wang, W. Good","doi":"10.1109/IJCNN.1999.830835","DOIUrl":"https://doi.org/10.1109/IJCNN.1999.830835","url":null,"abstract":"Artificial neural networks (ANN) have been widely used in computer-assisted diagnosis (CAD) schemes as a classification tool to identify abnormalities in digitized mammograms. Because of certain limitations of ANNs, some investigators argue that Bayesian belief network (BBN) may exhibit higher performance. In this study we compared the performance of an ANN and a BBN used in the same CAD scheme. The common databases and the same genetic algorithm (GA) were used to optimize both networks. The experimental results demonstrated that using GA optimization, the performance of the two networks converged to the same level in detecting masses from digitized mammograms. Therefore, in this study we concluded that improving the performance of CAD schemes might be more dependent on optimization of feature selection and diversity of training database than on any particular machine classification paradigm.","PeriodicalId":157719,"journal":{"name":"IJCNN'99. International Joint Conference on Neural Networks. Proceedings (Cat. No.99CH36339)","volume":"117 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132842384","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1999-07-10DOI: 10.1109/IJCNN.1999.836194
M. S. Zamani, Farhad Mehdipour
In this paper a Kohonen map-based algorithm for the placement of gate arrays and standard cells is presented. An abstract specification of the design is converted to a set of appropriate input vectors using a mathematical method, called "multidimensional scaling". These vectors which have, in general, higher dimensionality are fed to the self-organizing map at random in order to map them onto a 2D plane of the regular chip. The mapping is done in such a way that the cells with higher connectivity are placed close to each other, hence minimizing total connection length in the design. Two processes, called reassignment and rearrangement, are employed to make the algorithm applicable to the standard cell designs. In addition to the small examples introduced in other papers, two standard cell benchmarks were tried and better results were observed for these large designs compared to other neural net-barred approaches.
{"title":"An efficient method for placement of VLSI designs with Kohonen map","authors":"M. S. Zamani, Farhad Mehdipour","doi":"10.1109/IJCNN.1999.836194","DOIUrl":"https://doi.org/10.1109/IJCNN.1999.836194","url":null,"abstract":"In this paper a Kohonen map-based algorithm for the placement of gate arrays and standard cells is presented. An abstract specification of the design is converted to a set of appropriate input vectors using a mathematical method, called \"multidimensional scaling\". These vectors which have, in general, higher dimensionality are fed to the self-organizing map at random in order to map them onto a 2D plane of the regular chip. The mapping is done in such a way that the cells with higher connectivity are placed close to each other, hence minimizing total connection length in the design. Two processes, called reassignment and rearrangement, are employed to make the algorithm applicable to the standard cell designs. In addition to the small examples introduced in other papers, two standard cell benchmarks were tried and better results were observed for these large designs compared to other neural net-barred approaches.","PeriodicalId":157719,"journal":{"name":"IJCNN'99. International Joint Conference on Neural Networks. Proceedings (Cat. No.99CH36339)","volume":"22 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133565754","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1999-07-10DOI: 10.1109/IJCNN.1999.836188
Leenesh Kumar Maisuria, Cheng Soon Ong, W. Lai
Password authentication is the most commonly used identification system in today's computer world. Its security can be enhanced using typing biometrics as a transparent layer of user authentication. Our research focuses on using the time period between keystrokes as the measure of the individual's typing pattern. The typing pattern of a particular individual can be represented by the weights of a fully trained multilayer perceptron. Alternatively, each user's typing pattern can be viewed as a cluster of measurements that can be differentiated from clusters of other users.
{"title":"A comparison of artificial neural networks and cluster analysis for typing biometrics authentication","authors":"Leenesh Kumar Maisuria, Cheng Soon Ong, W. Lai","doi":"10.1109/IJCNN.1999.836188","DOIUrl":"https://doi.org/10.1109/IJCNN.1999.836188","url":null,"abstract":"Password authentication is the most commonly used identification system in today's computer world. Its security can be enhanced using typing biometrics as a transparent layer of user authentication. Our research focuses on using the time period between keystrokes as the measure of the individual's typing pattern. The typing pattern of a particular individual can be represented by the weights of a fully trained multilayer perceptron. Alternatively, each user's typing pattern can be viewed as a cluster of measurements that can be differentiated from clusters of other users.","PeriodicalId":157719,"journal":{"name":"IJCNN'99. International Joint Conference on Neural Networks. Proceedings (Cat. No.99CH36339)","volume":"35 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133273923","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1999-07-10DOI: 10.1109/IJCNN.1999.832729
Gongzhan Wang, Tzong-Jing Lee
In this article a new neural-network based cross-coupled control algorithm that integrates the cross-coupled control and neural network techniques together is presented In this neural network based cross-coupled control system, fixed gain PID controller for each individual axis is replaced by a heuristic neural network learning controller. The conventional cross-coupled controller is substituted by an efficient neural network cross-coupled controller. Experimental results show that the proposed new neural network based cross-coupled control scheme can be successfully applied to the precise circular tracking problem of a nonlinear uncertain linear motor X-Y table. It is also demonstrated that performance of the neural network based cross-coupled control scheme is superior to the conventional cross-coupled control scheme.
{"title":"Neural-network cross-coupled control system with application on circular tracking of linear motor X-Y table","authors":"Gongzhan Wang, Tzong-Jing Lee","doi":"10.1109/IJCNN.1999.832729","DOIUrl":"https://doi.org/10.1109/IJCNN.1999.832729","url":null,"abstract":"In this article a new neural-network based cross-coupled control algorithm that integrates the cross-coupled control and neural network techniques together is presented In this neural network based cross-coupled control system, fixed gain PID controller for each individual axis is replaced by a heuristic neural network learning controller. The conventional cross-coupled controller is substituted by an efficient neural network cross-coupled controller. Experimental results show that the proposed new neural network based cross-coupled control scheme can be successfully applied to the precise circular tracking problem of a nonlinear uncertain linear motor X-Y table. It is also demonstrated that performance of the neural network based cross-coupled control scheme is superior to the conventional cross-coupled control scheme.","PeriodicalId":157719,"journal":{"name":"IJCNN'99. International Joint Conference on Neural Networks. Proceedings (Cat. No.99CH36339)","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133450717","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：