Pub Date : 1999-11-16DOI: 10.1109/ICONIP.1999.843989
M. Watts
The paper presents an investigation into the properties of evolving fuzzy neural networks. It is shown that for the task of isolated phoneme recognition these networks are resistant to forgetting, highly adaptive and possess good generalisation capabilities. It is also shown which training parameters are most relevant to the behaviour of the network, and what effect adjustment of these parameters will have.
{"title":"An investigation of the properties of evolving fuzzy neural networks","authors":"M. Watts","doi":"10.1109/ICONIP.1999.843989","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.843989","url":null,"abstract":"The paper presents an investigation into the properties of evolving fuzzy neural networks. It is shown that for the task of isolated phoneme recognition these networks are resistant to forgetting, highly adaptive and possess good generalisation capabilities. It is also shown which training parameters are most relevant to the behaviour of the network, and what effect adjustment of these parameters will have.","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"130 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122113882","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 : 1999-11-16DOI: 10.1109/ICONIP.1999.844697
A. Adams, J. Rienks
A backpropagation neural network has been used as a determining tool to extract relationships and qualitative information from a small database on the microhabitats and colour patterns of male and female jumping spiders.
{"title":"Using a neural network to extract biological information from a jumping spider database","authors":"A. Adams, J. Rienks","doi":"10.1109/ICONIP.1999.844697","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.844697","url":null,"abstract":"A backpropagation neural network has been used as a determining tool to extract relationships and qualitative information from a small database on the microhabitats and colour patterns of male and female jumping spiders.","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"84 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122221383","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 : 1999-11-16DOI: 10.1109/ICONIP.1999.845660
J. McCullagh, K. Bluff, T. Hendtlass
Various techniques for estimating meteorological parameters have been developed over the past few years that involve artificial neural networks. However, the estimation of rainfall has continued to be a very difficult and complex problem to solve. Data mining techniques are needed to extract the important information from the vast amount of meteorological data available. A single multi-layer backpropagation neural network used on complex problems involving different sub-tasks will often show strong inter sub-task interference effects that lead to slow learning and poor generalisation. Dividing the system up into several different "expert networks" each specialising in a different sub-task can reduce this interference at the cost of having to combine the outputs from each of the experts. This paper investigates the technique of dividing the rainfall estimation problem into a number of such experts each specialising in a particular rainfall band (i.e. low, medium or high rain). Results demonstrate that expert networks can be successfully developed which result in both improved individual classifications and improved overall classification accuracy.
{"title":"Evolving expert neural networks for meteorological rainfall estimations","authors":"J. McCullagh, K. Bluff, T. Hendtlass","doi":"10.1109/ICONIP.1999.845660","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.845660","url":null,"abstract":"Various techniques for estimating meteorological parameters have been developed over the past few years that involve artificial neural networks. However, the estimation of rainfall has continued to be a very difficult and complex problem to solve. Data mining techniques are needed to extract the important information from the vast amount of meteorological data available. A single multi-layer backpropagation neural network used on complex problems involving different sub-tasks will often show strong inter sub-task interference effects that lead to slow learning and poor generalisation. Dividing the system up into several different \"expert networks\" each specialising in a different sub-task can reduce this interference at the cost of having to combine the outputs from each of the experts. This paper investigates the technique of dividing the rainfall estimation problem into a number of such experts each specialising in a particular rainfall band (i.e. low, medium or high rain). Results demonstrate that expert networks can be successfully developed which result in both improved individual classifications and improved overall classification accuracy.","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"31 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123335751","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 : 1999-11-16DOI: 10.1109/ICONIP.1999.843973
M. Yasunaga, T. Nakamura, I. Yoshihara
We have developed a new design methodology for face identification chips using a genetic algorithm. In the design, face images are transformed to truth-tables and they are evolved to obtain generalization ability. Digital circuits are synthesized by using the evolved truth-tables. Parallelism in the data can be embedded in the circuits by this direct hardware implementation of the face images. A face identification chip prototype has been developed by synthesizing the evolved truth tables to logic circuits. The circuit size of the chip was 1334 gates for one person on average, and this was small enough to be implemented onto a standard FPGA (field programmable gate array) chip. The chip identified a face image at 400 ns and achieved an identification accuracy of 97.2% in average. Furthermore, a high identification accuracy of more than 90% was maintained even under 18% faulty gate ratio and this high fault tolerance degraded gracefully as the faulty gate ratio increased.
{"title":"A fault-tolerant evolvable face identification chip","authors":"M. Yasunaga, T. Nakamura, I. Yoshihara","doi":"10.1109/ICONIP.1999.843973","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.843973","url":null,"abstract":"We have developed a new design methodology for face identification chips using a genetic algorithm. In the design, face images are transformed to truth-tables and they are evolved to obtain generalization ability. Digital circuits are synthesized by using the evolved truth-tables. Parallelism in the data can be embedded in the circuits by this direct hardware implementation of the face images. A face identification chip prototype has been developed by synthesizing the evolved truth tables to logic circuits. The circuit size of the chip was 1334 gates for one person on average, and this was small enough to be implemented onto a standard FPGA (field programmable gate array) chip. The chip identified a face image at 400 ns and achieved an identification accuracy of 97.2% in average. Furthermore, a high identification accuracy of more than 90% was maintained even under 18% faulty gate ratio and this high fault tolerance degraded gracefully as the faulty gate ratio increased.","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"19 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131903710","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 : 1999-11-16DOI: 10.1109/ICONIP.1999.845695
M. Towsey, J. Diederich
Neural models of cortical function frequently assume initial profuse connectivity and ignore issues of cortical development. There is increasing interest in cortical models that minimise pre-specification of architecture and instead allow input and learning rules to sculpt connectivity. We describe a model of cortical development that begins with minimal connectivity but arrives at useful functionality through a variety of mechanisms, including Hebbian learning, volume learning, synaptic sprouting and structured input. We discuss some of the issues pertinent to the building of neural structure.
{"title":"Modeling cortical function starting with minimal connectivity","authors":"M. Towsey, J. Diederich","doi":"10.1109/ICONIP.1999.845695","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.845695","url":null,"abstract":"Neural models of cortical function frequently assume initial profuse connectivity and ignore issues of cortical development. There is increasing interest in cortical models that minimise pre-specification of architecture and instead allow input and learning rules to sculpt connectivity. We describe a model of cortical development that begins with minimal connectivity but arrives at useful functionality through a variety of mechanisms, including Hebbian learning, volume learning, synaptic sprouting and structured input. We discuss some of the issues pertinent to the building of neural structure.","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"6 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131951893","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 : 1999-11-16DOI: 10.1109/ICONIP.1999.845645
M. Sasaki, M. Kawafuku, K. Takahashi
Both neural networks and immunity-based systems are biologically inspired techniques that have the capability of identifying and controlling. The information processing principles of these natural systems inspired the development of intelligent problem solving techniques, namely, the artificial neural network and the artificial immune system. An adaptive learning method for a neural network (NN) controller using an immune feedback law is proposed. The immune feedback law features rapid response to foreign matter and rapid stabilization of biological immune systems. Several improvements can be made to improve gradient descent NN learning algorithms. The use of an adaptive learning rate attempts to keep the learning step size as large as possible while keeping learning stable. In the proposed method, because the immune feedback law changes the learning rate of the NN individually and adaptively, it is expected that a cost function is rapidly minimized and learning time is decreased. In the control structure, a reference signal self-organizing control system using NNs for flexible microactuators is used. In this system, the NN functions as a reference input filter, setting new reference signals in the closed loop system. Numerical and experimental results show that the proposed control system is effective in tracking a reference signal.
{"title":"An immune feedback mechanism based adaptive learning of neural network controller","authors":"M. Sasaki, M. Kawafuku, K. Takahashi","doi":"10.1109/ICONIP.1999.845645","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.845645","url":null,"abstract":"Both neural networks and immunity-based systems are biologically inspired techniques that have the capability of identifying and controlling. The information processing principles of these natural systems inspired the development of intelligent problem solving techniques, namely, the artificial neural network and the artificial immune system. An adaptive learning method for a neural network (NN) controller using an immune feedback law is proposed. The immune feedback law features rapid response to foreign matter and rapid stabilization of biological immune systems. Several improvements can be made to improve gradient descent NN learning algorithms. The use of an adaptive learning rate attempts to keep the learning step size as large as possible while keeping learning stable. In the proposed method, because the immune feedback law changes the learning rate of the NN individually and adaptively, it is expected that a cost function is rapidly minimized and learning time is decreased. In the control structure, a reference signal self-organizing control system using NNs for flexible microactuators is used. In this system, the NN functions as a reference input filter, setting new reference signals in the closed loop system. Numerical and experimental results show that the proposed control system is effective in tracking a reference signal.","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"2000 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129543133","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 : 1999-11-16DOI: 10.1109/ICONIP.1999.844688
S. Mcgibney, A. Zaknich
A multilayer perceptron classifier is applied to the classification of gas flow states. A number of suitable discriminate features are determined heuristically for the categorization of gas flow states, including the background (machinery and wind tunnel noise), laminar flow (sinusoidal signal), transition 1 (frequency-resonant shifts), transition 2 (instantaneous changes in phase and turbulent characteristics) and turbulent flow (random noise). This technique can be used to develop an automatic real-time classifier for gas flow.
{"title":"Unsteady airflow classification by artificial neural networks","authors":"S. Mcgibney, A. Zaknich","doi":"10.1109/ICONIP.1999.844688","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.844688","url":null,"abstract":"A multilayer perceptron classifier is applied to the classification of gas flow states. A number of suitable discriminate features are determined heuristically for the categorization of gas flow states, including the background (machinery and wind tunnel noise), laminar flow (sinusoidal signal), transition 1 (frequency-resonant shifts), transition 2 (instantaneous changes in phase and turbulent characteristics) and turbulent flow (random noise). This technique can be used to develop an automatic real-time classifier for gas flow.","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129546515","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 : 1999-11-16DOI: 10.1109/ICONIP.1999.843966
M. Flax, J.S. Jin
Bidirectional systems are used to transform between known or desired input and output data in either direction. This paper compares two different methods for transforming from output to input data. It is outlined for transformation systems where the forward transform uses a kernel which has been adaptively/iteratively found but require an inversion scheme which maps exactly the forward transformation.
{"title":"A matrix approach to neural network inversion","authors":"M. Flax, J.S. Jin","doi":"10.1109/ICONIP.1999.843966","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.843966","url":null,"abstract":"Bidirectional systems are used to transform between known or desired input and output data in either direction. This paper compares two different methods for transforming from output to input data. It is outlined for transformation systems where the forward transform uses a kernel which has been adaptively/iteratively found but require an inversion scheme which maps exactly the forward transformation.","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"28 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131223154","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 : 1999-11-16DOI: 10.1109/ICONIP.1999.845668
A. Kaplan
A critical review of the principal strategies of the EEG description as a piecewise stationary process is given and new methodology of EEG segmentation, based on nonparametric statistical analysis, is proposed. Our methodology provides the detection of moments of quasi-stationary segments' boundaries in almost any EEG characteristic for a given level of false alarm probability. Relatively high temporal resolution of the method makes it possible to formulate a new approach to investigation of the functional synchrony between different brain areas. We discuss also the achievements, problems, and prospects of EEG signal segmentation.
{"title":"Segmental structure of EEG more likely reveals the dynamic multistability of the brain tissue than the continual plasticity one","authors":"A. Kaplan","doi":"10.1109/ICONIP.1999.845668","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.845668","url":null,"abstract":"A critical review of the principal strategies of the EEG description as a piecewise stationary process is given and new methodology of EEG segmentation, based on nonparametric statistical analysis, is proposed. Our methodology provides the detection of moments of quasi-stationary segments' boundaries in almost any EEG characteristic for a given level of false alarm probability. Relatively high temporal resolution of the method makes it possible to formulate a new approach to investigation of the functional synchrony between different brain areas. We discuss also the achievements, problems, and prospects of EEG signal segmentation.","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133856224","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 : 1999-11-16DOI: 10.1109/ICONIP.1999.845647
G. Holmes, L. Trigg
The process of developing applications of machine learning and data mining that employ supervised classification algorithms includes the important step of knowledge verification. Interpretable output is presented to a user so that they can verify that the knowledge contained in the output makes sense for the given application. As the development of an application is an iterative process it is quite likely that a user would wish to compare models constructed at various times or stages. One crucial stage where comparison of models is important is when the accuracy of a model is being estimated, typically using some form of cross-validation. This stage is used to establish an estimate of how well a model will perform on unseen data. This is vital information to present to a user, but it is also important to show the degree of variation between models obtained from the entire dataset and models obtained during cross-validation. In this way it can be verified that the cross-validation models are at least structurally aligned with the model garnered from the entire dataset. This paper presents a diagnostic tool for the comparison of tree-based supervised classification models. The method is adapted from work on approximate tree matching and applied to decision trees. The tool is described together with experimental results on standard datasets.
{"title":"A diagnostic tool for tree based supervised classification learning algorithms","authors":"G. Holmes, L. Trigg","doi":"10.1109/ICONIP.1999.845647","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.845647","url":null,"abstract":"The process of developing applications of machine learning and data mining that employ supervised classification algorithms includes the important step of knowledge verification. Interpretable output is presented to a user so that they can verify that the knowledge contained in the output makes sense for the given application. As the development of an application is an iterative process it is quite likely that a user would wish to compare models constructed at various times or stages. One crucial stage where comparison of models is important is when the accuracy of a model is being estimated, typically using some form of cross-validation. This stage is used to establish an estimate of how well a model will perform on unseen data. This is vital information to present to a user, but it is also important to show the degree of variation between models obtained from the entire dataset and models obtained during cross-validation. In this way it can be verified that the cross-validation models are at least structurally aligned with the model garnered from the entire dataset. This paper presents a diagnostic tool for the comparison of tree-based supervised classification models. The method is adapted from work on approximate tree matching and applied to decision trees. The tool is described together with experimental results on standard datasets.","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"199 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134167525","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：