Pub Date : 1999-11-16DOI: 10.1109/ICONIP.1999.843961
T. Kindo, H. Yoshida, M. Hirahara
A personal information categorizing system, PICSY, on an adaptive personal information filtering system is presented. PICSY is an information catergorizing system based on an associative memory model. Information categorization systems are needed because the amount of available information has become huge. PICSY is a practical information categorization system that divides each personal profile into sub-profiles which respectively correspond to categories of user's interests. Our field test shows that PICSY extracts several categories from each personal profile. The extracted categories are reasonable for users to recognize their interests. The results of the field test support that PICSY can be put to practical use.
{"title":"Personal information categorizing system with an associative memory model","authors":"T. Kindo, H. Yoshida, M. Hirahara","doi":"10.1109/ICONIP.1999.843961","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.843961","url":null,"abstract":"A personal information categorizing system, PICSY, on an adaptive personal information filtering system is presented. PICSY is an information catergorizing system based on an associative memory model. Information categorization systems are needed because the amount of available information has become huge. PICSY is a practical information categorization system that divides each personal profile into sub-profiles which respectively correspond to categories of user's interests. Our field test shows that PICSY extracts several categories from each personal profile. The extracted categories are reasonable for users to recognize their interests. The results of the field test support that PICSY can be put to practical use.","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"62 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":"130948504","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.1999.843991
A. Hirabayashi, H. Ogawa
When we are concerned with a learning method, such as regularization learning, which does not directly deal with generalization error, we usually use it to achieve some "true objective learning". That is, we first have some objective such as minimization of generalization error, then we look for a learning method which could achieve the objective learning. There is, however, another situation. When we have developed a learning method, we wish to apply it to a wide range of different purposes. We discuss the latter problem. We clarify the bound of applicability of memorization learning within a family of projection learning. The bound is determined by the location of sample points and the nature of noise.
{"title":"What can memorization learning do from noisy training examples?","authors":"A. Hirabayashi, H. Ogawa","doi":"10.1109/ICONIP.1999.843991","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.843991","url":null,"abstract":"When we are concerned with a learning method, such as regularization learning, which does not directly deal with generalization error, we usually use it to achieve some \"true objective learning\". That is, we first have some objective such as minimization of generalization error, then we look for a learning method which could achieve the objective learning. There is, however, another situation. When we have developed a learning method, we wish to apply it to a wide range of different purposes. We discuss the latter problem. We clarify the bound of applicability of memorization learning within a family of projection learning. The bound is determined by the location of sample points and the nature of noise.","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":"133897973","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.1999.844651
X.Q. Li, I. King
Efficient and accurate information retrieval (IR) is one of the main issues in multimedia databases. Clustering can help to generate the efficient indexing structures and provide the comparison between data types. The Most Expressive Feature (MEF) extraction can improve comparison accuracy between two data which belong to a same data type since it discards redundant features. The authors introduce a local linear principal component analysis (LLPCA) to design an optimal scheme for IR. The LLPCA realizes the clustering and local MEF extraction at the same time. Using these clusters and local MEFs, an IR scheme can be divided into two steps from coarse to fine. We apply the scheme to a trademark retrieval system to evaluate its performance based on the accuracy and efficiency measurements. The experimental results indicate this retrieval scheme is superior the other schemes using the original features or global MEFs extracted by a Global Linear PCA (GLPCA).
{"title":"Information retrieval using local linear PCA","authors":"X.Q. Li, I. King","doi":"10.1109/ICONIP.1999.844651","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.844651","url":null,"abstract":"Efficient and accurate information retrieval (IR) is one of the main issues in multimedia databases. Clustering can help to generate the efficient indexing structures and provide the comparison between data types. The Most Expressive Feature (MEF) extraction can improve comparison accuracy between two data which belong to a same data type since it discards redundant features. The authors introduce a local linear principal component analysis (LLPCA) to design an optimal scheme for IR. The LLPCA realizes the clustering and local MEF extraction at the same time. Using these clusters and local MEFs, an IR scheme can be divided into two steps from coarse to fine. We apply the scheme to a trademark retrieval system to evaluate its performance based on the accuracy and efficiency measurements. The experimental results indicate this retrieval scheme is superior the other schemes using the original features or global MEFs extracted by a Global Linear PCA (GLPCA).","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"96 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":"129698567","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.1999.845674
S. Yamaguchi, H. Itakura
A new modular neural network architecture and its learning algorithm are proposed for a mobile robot controller. The learning algorithm for the proposed new network architecture is based on a feedback error learning procedure, which requires a feedback controller for training processes. It is not so easy, however, to obtain a robot feedback controller, when the robot control task is much more complex. In the present architecture, the complex robot control task is divided into a couple of small simple tasks, each of which is assigned to each of small network modules, respectively. By dividing the complex task, the simple feedback controllers are assigned to the network modules. Therefore, the neural network in each module can be trained by the feedback error learning scheme. The command to the robots is the weighted sum of the outputs of the modules. The weights for each module are obtained from a neural network which is one of the network modules in our proposed architecture. The present neural network architecture and learning algorithm are applied to a set of several robot controllers, whose task is to push a large box to a goal. It is confirmed through computer simulation experiments that the algorithm can train the robot controller skillfully.
{"title":"A modular neural network for control of mobile robots","authors":"S. Yamaguchi, H. Itakura","doi":"10.1109/ICONIP.1999.845674","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.845674","url":null,"abstract":"A new modular neural network architecture and its learning algorithm are proposed for a mobile robot controller. The learning algorithm for the proposed new network architecture is based on a feedback error learning procedure, which requires a feedback controller for training processes. It is not so easy, however, to obtain a robot feedback controller, when the robot control task is much more complex. In the present architecture, the complex robot control task is divided into a couple of small simple tasks, each of which is assigned to each of small network modules, respectively. By dividing the complex task, the simple feedback controllers are assigned to the network modules. Therefore, the neural network in each module can be trained by the feedback error learning scheme. The command to the robots is the weighted sum of the outputs of the modules. The weights for each module are obtained from a neural network which is one of the network modules in our proposed architecture. The present neural network architecture and learning algorithm are applied to a set of several robot controllers, whose task is to push a large box to a goal. It is confirmed through computer simulation experiments that the algorithm can train the robot controller skillfully.","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"14 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":"130052095","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.1999.844696
V.K. Sagai, A. Koh Jit Beng
This paper investigates and proposes the fusion of fuzzy logic and neural network technology in automated fingerprint recognition for the extraction of important fingerprint features, also known as minutiae.
本文研究并提出了模糊逻辑和神经网络技术在自动指纹识别中的融合，以提取指纹的重要特征，也称为细节。
{"title":"Fingerprint feature extraction by fuzzy logic and neural networks","authors":"V.K. Sagai, A. Koh Jit Beng","doi":"10.1109/ICONIP.1999.844696","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.844696","url":null,"abstract":"This paper investigates and proposes the fusion of fuzzy logic and neural network technology in automated fingerprint recognition for the extraction of important fingerprint features, also known as minutiae.","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":"128219295","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.1999.844014
I.I. Esat, B. Kothari, A. Shaikh, P. Wrathall
Investigates the direct encoding scheme in a neural network representation in the context of network construction using a genetic algorithm (GA). This paper addresses the use and the success of direct encoding schemes, in particular a specific scheme previously proposed by B.C. Kothari and I.I. Esat (1st World Conf. in Integrated Design and Process Technol., pp. 234-45, 1995). An investigation shows that obtaining the results previously presented by Kothari has not been possible, and the very high success reported has not been verified. However, the implementation reported in this paper does produce modular networks with improved training, as previously reported.
{"title":"Encoding neural networks for GA based structural construction","authors":"I.I. Esat, B. Kothari, A. Shaikh, P. Wrathall","doi":"10.1109/ICONIP.1999.844014","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.844014","url":null,"abstract":"Investigates the direct encoding scheme in a neural network representation in the context of network construction using a genetic algorithm (GA). This paper addresses the use and the success of direct encoding schemes, in particular a specific scheme previously proposed by B.C. Kothari and I.I. Esat (1st World Conf. in Integrated Design and Process Technol., pp. 234-45, 1995). An investigation shows that obtaining the results previously presented by Kothari has not been possible, and the very high success reported has not been verified. However, the implementation reported in this paper does produce modular networks with improved training, as previously reported.","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"55 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":"134182595","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.1999.844680
R. Romero, J. Kacprzyk, F. Gomide
A neural network with a two-layer feedback topology and generalized recurrent neurons, for solving nonlinear discrete dynamic optimization problems is developed. The algorithm is based on R. Bellmann's (1957) optimality principle and the interchange of information during synaptic chemical processing among neurons. The technique is applied to solve fuzzy decision making problems.
{"title":"A biologically inspired neural network for dynamic system optimization","authors":"R. Romero, J. Kacprzyk, F. Gomide","doi":"10.1109/ICONIP.1999.844680","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.844680","url":null,"abstract":"A neural network with a two-layer feedback topology and generalized recurrent neurons, for solving nonlinear discrete dynamic optimization problems is developed. The algorithm is based on R. Bellmann's (1957) optimality principle and the interchange of information during synaptic chemical processing among neurons. The technique is applied to solve fuzzy decision making 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":"37 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":"132862972","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.1999.843952
I. Aleksander, B. Dunmall, V.D. Frate
How can a system with visual input become capable of visualising what is meant by new combinations of known words? For example, it is possible for most of us to visualise a blue banana with red spots even though such an object would never have formed part of our experience. The authors discuss a neural system which is capable of simple forms of this kind of visualisation. It is shown that success in this task depends on the activity of a neural module whose firing patterns represent the 'visual awareness' of the system and the way that this module interacts with others in the system. The paper discloses the first set of results from this ongoing research project.
{"title":"Recent progress on neural models of seeing and visualisation","authors":"I. Aleksander, B. Dunmall, V.D. Frate","doi":"10.1109/ICONIP.1999.843952","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.843952","url":null,"abstract":"How can a system with visual input become capable of visualising what is meant by new combinations of known words? For example, it is possible for most of us to visualise a blue banana with red spots even though such an object would never have formed part of our experience. The authors discuss a neural system which is capable of simple forms of this kind of visualisation. It is shown that success in this task depends on the activity of a neural module whose firing patterns represent the 'visual awareness' of the system and the way that this module interacts with others in the system. The paper discloses the first set of results from this ongoing research project.","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"34 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":"130547213","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.1999.845688
M. Ishikawa, H. Matsumura
We have succeeded in recognizing 32 kinds of hand shapes based on self-organization by measuring the angles of 10 joints of a hand using a DataGlove. Recognition of hand gestures, however, is far more difficult, because it must recognize a sequence of hand shapes instead of its snapshot. An essential difficulty in gesture recognition is how to deal with a sequence of body postures or hand shapes. Since a hand shape is represented by a 10-dimensional (10D) vector measured by a DataGlove, a hand gesture is represented by a sequence of 10D vectors. Our proposal is to recognize a hand gesture by the following procedure. (1) Angles of finger joints are measured at some time interval by a DataGlove. (2) Each gesture is segmented from a sequence of hand shapes. (3) The data length, i.e. the number of snapshots in each gesture, is adjusted to obtain data of a fixed length. (4) An input vector for self-organization is obtained by connecting a sequence of 10D hand-shape vectors. (5) Clustering of hand gestures is carried out by self-organization according to their similarities. Since self-organization is not directly applicable to time series data, the fourth step is the key idea for recognition. We present a detailed description of the proposed recognition method. We then give an overview of hand-gesture data obtained by a DataGlove. Finally, the results of some recognition experiments are provided. This is followed by discussions and conclusions.
{"title":"Recognition of a hand-gesture based on self-organization using a DataGlove","authors":"M. Ishikawa, H. Matsumura","doi":"10.1109/ICONIP.1999.845688","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.845688","url":null,"abstract":"We have succeeded in recognizing 32 kinds of hand shapes based on self-organization by measuring the angles of 10 joints of a hand using a DataGlove. Recognition of hand gestures, however, is far more difficult, because it must recognize a sequence of hand shapes instead of its snapshot. An essential difficulty in gesture recognition is how to deal with a sequence of body postures or hand shapes. Since a hand shape is represented by a 10-dimensional (10D) vector measured by a DataGlove, a hand gesture is represented by a sequence of 10D vectors. Our proposal is to recognize a hand gesture by the following procedure. (1) Angles of finger joints are measured at some time interval by a DataGlove. (2) Each gesture is segmented from a sequence of hand shapes. (3) The data length, i.e. the number of snapshots in each gesture, is adjusted to obtain data of a fixed length. (4) An input vector for self-organization is obtained by connecting a sequence of 10D hand-shape vectors. (5) Clustering of hand gestures is carried out by self-organization according to their similarities. Since self-organization is not directly applicable to time series data, the fourth step is the key idea for recognition. We present a detailed description of the proposed recognition method. We then give an overview of hand-gesture data obtained by a DataGlove. Finally, the results of some recognition experiments are provided. This is followed by discussions and conclusions.","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"40 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":"125194980","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.1999.845643
S. Sugiyama
Lots of work has been done in the field of AI, knowledge bases, natural languages, semantics and logic by using the tree search method, pattern recognition, neural networks, etc., and we are now beginning to have systems which can understand the meaning of a word or a sentence as a human does, but these systems are not flexible or mature enough for real use, and so not yet applicable for real use. This problem mainly comes from the processing methods used, the methods used to understand words and sentences, and the non-dynamic recognition behaviours. So, in this paper, I introduce a semantic and logic processing method, using neural networks, which has a unique way of transforming words and sentences into neural networks and dynamical behaviourism-accomplishing objectives. As a result of these processes, I found that a sentence has a meaning that is related to certain knowledge, and this sentence-to-knowledge transformation has a unique knowledge compression method. Therefore, I also introduce a knowledge compression method in semantics and logic.
{"title":"Simple pre-processor for semantics and logic","authors":"S. Sugiyama","doi":"10.1109/ICONIP.1999.845643","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.845643","url":null,"abstract":"Lots of work has been done in the field of AI, knowledge bases, natural languages, semantics and logic by using the tree search method, pattern recognition, neural networks, etc., and we are now beginning to have systems which can understand the meaning of a word or a sentence as a human does, but these systems are not flexible or mature enough for real use, and so not yet applicable for real use. This problem mainly comes from the processing methods used, the methods used to understand words and sentences, and the non-dynamic recognition behaviours. So, in this paper, I introduce a semantic and logic processing method, using neural networks, which has a unique way of transforming words and sentences into neural networks and dynamical behaviourism-accomplishing objectives. As a result of these processes, I found that a sentence has a meaning that is related to certain knowledge, and this sentence-to-knowledge transformation has a unique knowledge compression method. Therefore, I also introduce a knowledge compression method in semantics and logic.","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"32 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":"127089964","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：