Pub Date : 1999-11-16DOI: 10.1109/ICONIP.1999.845687
C. Fung, Kok Wai Wong, D. Myers
A neural network-based data analysis model for the prediction and classification of field data has many attractions. However, there are problems in ensuring the generalisation capability of the data analysis model, in measuring the similarity between the original training data and the new unknown data and in processing large data volumes. This paper reports the use of self-organising maps (SOM) to overcome these difficulties and illustrates the utilisation of this approach though applications in the agricultural, resource exploration and mineral processing areas.
{"title":"An intelligent data analysis approach using self-organising-maps","authors":"C. Fung, Kok Wai Wong, D. Myers","doi":"10.1109/ICONIP.1999.845687","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.845687","url":null,"abstract":"A neural network-based data analysis model for the prediction and classification of field data has many attractions. However, there are problems in ensuring the generalisation capability of the data analysis model, in measuring the similarity between the original training data and the new unknown data and in processing large data volumes. This paper reports the use of self-organising maps (SOM) to overcome these difficulties and illustrates the utilisation of this approach though applications in the agricultural, resource exploration and mineral processing areas.","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"24 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":"126617458","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.1999.845663
V. Brusic, John Zeleznikow, T. Sturniolo, E. Bono, J. Hammer
Knowledge discovery from databases (KDD) in biology largely depends on the use of accurate computer models of biological processes. KDD applications in immunology include the discovery of vaccine targets and new functional relations within the immune system. We describe a process of development and refinement of artificial neural network models of the human HLA-DR1 molecule, useful for the discovery of peptide vaccines. High accuracy of these models was achieved by data cleansing techniques and by cyclical retraining using new data.
{"title":"Data cleansing for computer models: a case study from immunology","authors":"V. Brusic, John Zeleznikow, T. Sturniolo, E. Bono, J. Hammer","doi":"10.1109/ICONIP.1999.845663","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.845663","url":null,"abstract":"Knowledge discovery from databases (KDD) in biology largely depends on the use of accurate computer models of biological processes. KDD applications in immunology include the discovery of vaccine targets and new functional relations within the immune system. We describe a process of development and refinement of artificial neural network models of the human HLA-DR1 molecule, useful for the discovery of peptide vaccines. High accuracy of these models was achieved by data cleansing techniques and by cyclical retraining using new data.","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"29 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":"126805447","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.1999.843955
P. Munro, G. Hernández
A single framework is shown to encompass several existing learning rules, separating them into positive and negative terms, respectively corresponding to long-term potentiation (LTP) and long-term depression (LTD) phenomena. Each term is expressed as an integral of a Hebbian product over time, modulated by a kernel function. Carefully chosen kernel functions are shown to exhibit computational properties of temporal contrast enhancement and prediction. Some preliminary simulation results are presented for illustration purposes.
{"title":"An LTP/LTD perspective on learning rules","authors":"P. Munro, G. Hernández","doi":"10.1109/ICONIP.1999.843955","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.843955","url":null,"abstract":"A single framework is shown to encompass several existing learning rules, separating them into positive and negative terms, respectively corresponding to long-term potentiation (LTP) and long-term depression (LTD) phenomena. Each term is expressed as an integral of a Hebbian product over time, modulated by a kernel function. Carefully chosen kernel functions are shown to exhibit computational properties of temporal contrast enhancement and prediction. Some preliminary simulation results are presented for illustration purposes.","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"33 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":"127002744","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.1999.844011
Y. Wada, M. Kawato
Proposes a computational handwriting model based on the optimization principle. The computational theory, the representation level and the hardware involved are the minimum commanded torque-change criterion, a set of via-points extracted from handwritten characters and a forward-inverse-relaxation neural network model, respectively. However, for via-point representation in the model, both timing and spatial information are needed. In this paper, we propose a new model in which the time passing through via-points is estimated by optimizing the criterion. The model is studied theoretically, and it is shown that the trajectory generated by the model is the same as the data obtained from human subjects in experiments.
{"title":"A computational model for arm trajectory formation by optimization of via-point time","authors":"Y. Wada, M. Kawato","doi":"10.1109/ICONIP.1999.844011","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.844011","url":null,"abstract":"Proposes a computational handwriting model based on the optimization principle. The computational theory, the representation level and the hardware involved are the minimum commanded torque-change criterion, a set of via-points extracted from handwritten characters and a forward-inverse-relaxation neural network model, respectively. However, for via-point representation in the model, both timing and spatial information are needed. In this paper, we propose a new model in which the time passing through via-points is estimated by optimizing the criterion. The model is studied theoretically, and it is shown that the trajectory generated by the model is the same as the data obtained from human subjects in experiments.","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"41 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":"126702926","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.1999.843976
D. Nauck, R. Kruse
Neuro-fuzzy classification approaches aim at creating fuzzy classification rules from data by using learning techniques derived from neural networks. NEFCLASS is able to learn fuzzy rules and fuzzy sets by simple heuristics. The aim of NEFCLASS is to quickly create interpretable fuzzy classifiers. Most neuro-fuzzy approaches can only deal with numerical attributes and cannot handle missing values. The authors present recent advances in the learning algorithms of NEFCLASS that address those problems.
{"title":"Learning in neuro-fuzzy systems with symbolic attributes and missing values","authors":"D. Nauck, R. Kruse","doi":"10.1109/ICONIP.1999.843976","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.843976","url":null,"abstract":"Neuro-fuzzy classification approaches aim at creating fuzzy classification rules from data by using learning techniques derived from neural networks. NEFCLASS is able to learn fuzzy rules and fuzzy sets by simple heuristics. The aim of NEFCLASS is to quickly create interpretable fuzzy classifiers. Most neuro-fuzzy approaches can only deal with numerical attributes and cannot handle missing values. The authors present recent advances in the learning algorithms of NEFCLASS that address those problems.","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"39 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":"127122193","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.1999.845672
M. T. Chao, Thomas Bräunl, Anthony Zaknich
The paper describes an indoor autonomous vision based obstacle avoidance robot system. The vision part of the system converts forward looking greyscale camera images into edge images using Canny edge detection. Both edge image and sonar ranging information is used as stimuli by the behaviours that make up the reactive part of the system. These behaviours all run concurrently and they couple perception to actions to generate motor responses. A priority based subsumption coordinator selects the most appropriate response to direct the robot away from obstacles.
{"title":"Visually-guided obstacle avoidance","authors":"M. T. Chao, Thomas Bräunl, Anthony Zaknich","doi":"10.1109/ICONIP.1999.845672","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.845672","url":null,"abstract":"The paper describes an indoor autonomous vision based obstacle avoidance robot system. The vision part of the system converts forward looking greyscale camera images into edge images using Canny edge detection. Both edge image and sonar ranging information is used as stimuli by the behaviours that make up the reactive part of the system. These behaviours all run concurrently and they couple perception to actions to generate motor responses. A priority based subsumption coordinator selects the most appropriate response to direct the robot away from obstacles.","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"16 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":"126566003","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.1999.845691
Ciprian-Daniel Neagu, M. Negoita, V. Palade
A unified approach for integrating explicit and implicit knowledge in connectionist expert systems is proposed. The explicit knowledge is represented by discrete fuzzy rules, which are directly mapped into an equivalent multi-purpose neural network based on the MAPI neuron (A.F. Rocha et al., 1992). The learning result is a refinement process of data sets, which is represented in a module (or combination of modules) of classical feedforward structures incorporating implicit fuzzy rules. The combination of explicit and implicit knowledge modules is viewed as an iterative process in knowledge acquisition and refinement.
提出了连接主义专家系统中显式和隐式知识集成的统一方法。显性知识由离散模糊规则表示，这些规则直接映射到基于MAPI神经元的等效多用途神经网络中(A.F. Rocha et al.， 1992)。学习结果是数据集的细化过程，这些数据集以包含隐式模糊规则的经典前馈结构的模块(或模块组合)表示。显性和隐性知识模块的结合被视为知识获取和提炼的迭代过程。
{"title":"Aspects of integration of explicit and implicit knowledge in connectionist expert systems","authors":"Ciprian-Daniel Neagu, M. Negoita, V. Palade","doi":"10.1109/ICONIP.1999.845691","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.845691","url":null,"abstract":"A unified approach for integrating explicit and implicit knowledge in connectionist expert systems is proposed. The explicit knowledge is represented by discrete fuzzy rules, which are directly mapped into an equivalent multi-purpose neural network based on the MAPI neuron (A.F. Rocha et al., 1992). The learning result is a refinement process of data sets, which is represented in a module (or combination of modules) of classical feedforward structures incorporating implicit fuzzy rules. The combination of explicit and implicit knowledge modules is viewed as an iterative process in knowledge acquisition and refinement.","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":"123532633","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.1999.844715
H.N. Cheung, A. Bouzerdoum, W. Newland
Investigates the dynamic range compression and contrast enhancement properties of shunting inhibition cellular neural networks (SICNN) used for colour image enhancement. First, the SICNN is formulated according to its structure and then the formulation is expressed in a digital format so that simulations can be performed. The resulting digital SICNN is then applied to a 1D ramp function to study its behaviour as compared to the logarithm of the function. Then, the SICNN is applied to colour images; the results show that, besides performing contrast enhancement, the SICNN also improves the colour constancy of the images as well as their sharpness.
{"title":"Properties of shunting inhibitory cellular neural networks for colour image enhancement","authors":"H.N. Cheung, A. Bouzerdoum, W. Newland","doi":"10.1109/ICONIP.1999.844715","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.844715","url":null,"abstract":"Investigates the dynamic range compression and contrast enhancement properties of shunting inhibition cellular neural networks (SICNN) used for colour image enhancement. First, the SICNN is formulated according to its structure and then the formulation is expressed in a digital format so that simulations can be performed. The resulting digital SICNN is then applied to a 1D ramp function to study its behaviour as compared to the logarithm of the function. Then, the SICNN is applied to colour images; the results show that, besides performing contrast enhancement, the SICNN also improves the colour constancy of the images as well as their sharpness.","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"36 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":"121672987","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.1999.845682
M. Fukumi, K. Nakaura, N. Akamatsu
A method of extracting rules from a rotation-invariant neural pattern recognition system formed using a genetic algorithm (GA) is presented. In particular, deterministic mutation (DM) is utilized to improve its convergence properties. It is performed on the basis of the result of neural network structure learning. DM can evolve chromosomes of individuals to increase their fitness functions in a deterministic manner. In this paper, coin data are used as inputs. The coins used are a Japanese 500-yen coin and a South Korean 500-won coin, which are very similar. GA is utilized to reduce the number of connection weights in the neural network. The network weights surviving after training represent rules to perform pattern classification for the coin data. The rules are then extracted from the network. Furthermore, the network has a procedure to substitute signum units for hidden sigmoid ones in examining its recognition accuracy. It enables us to easily extract rules. Simulation results show that this approach can generate a simple network structure and, as a result, simple rules for coin data classification.
{"title":"Rule generation from a rotation-invariant neural pattern recognition system","authors":"M. Fukumi, K. Nakaura, N. Akamatsu","doi":"10.1109/ICONIP.1999.845682","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.845682","url":null,"abstract":"A method of extracting rules from a rotation-invariant neural pattern recognition system formed using a genetic algorithm (GA) is presented. In particular, deterministic mutation (DM) is utilized to improve its convergence properties. It is performed on the basis of the result of neural network structure learning. DM can evolve chromosomes of individuals to increase their fitness functions in a deterministic manner. In this paper, coin data are used as inputs. The coins used are a Japanese 500-yen coin and a South Korean 500-won coin, which are very similar. GA is utilized to reduce the number of connection weights in the neural network. The network weights surviving after training represent rules to perform pattern classification for the coin data. The rules are then extracted from the network. Furthermore, the network has a procedure to substitute signum units for hidden sigmoid ones in examining its recognition accuracy. It enables us to easily extract rules. Simulation results show that this approach can generate a simple network structure and, as a result, simple rules for coin data classification.","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"123 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":"121495477","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.1999.843988
L.-Q. Zhang, A. Cichocki, S. Amari
We present a novel method-filter decomposition approach, for multichannel blind deconvolution of non-minimum phase systems. In earlier work we developed an efficient natural gradient algorithm for causal FIR filters. In this paper we further study the natural gradient method for noncausal filters. We decompose the doubly finite filters into a product of two filters, a noncausal FIR filter and a causal FIR filter. The natural gradient algorithm is employed to train the causal FIR filter, and a novel information backpropagation algorithm is developed for training the noncausal FIR filter. Simulations are given to illustrate the effectiveness and validity of the algorithm.
{"title":"Multichannel blind deconvolution of non-minimum phase systems using information backpropagation","authors":"L.-Q. Zhang, A. Cichocki, S. Amari","doi":"10.1109/ICONIP.1999.843988","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.843988","url":null,"abstract":"We present a novel method-filter decomposition approach, for multichannel blind deconvolution of non-minimum phase systems. In earlier work we developed an efficient natural gradient algorithm for causal FIR filters. In this paper we further study the natural gradient method for noncausal filters. We decompose the doubly finite filters into a product of two filters, a noncausal FIR filter and a causal FIR filter. The natural gradient algorithm is employed to train the causal FIR filter, and a novel information backpropagation algorithm is developed for training the noncausal FIR filter. Simulations are given to illustrate the effectiveness and validity of the algorithm.","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"249 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":"124748824","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：