Pub Date : 2002-11-07DOI: 10.1109/NNSP.2002.1030044
E. Corchado, J. Koetsier, D. MacDonald, C. Fyfe
We investigate an extension of Hebbian learning in a principal component analysis network which has been derived to be optimal for a specific probability density function(PDF). We note that this probability density function is one of a family of PDFs and investigate the learning rules formed in order to be optimal for several members of this family. We show that, whereas previous authors have viewed the single member of the family as an extension of PCA, it is more appropriate to view the whole family of learning rules as methods of performing exploratory projection pursuit (EPP). We explore the performance of our method first in response to an artificial data type, then to a real data set.
{"title":"Classification and ICA using maximum likelihood Hebbian learning","authors":"E. Corchado, J. Koetsier, D. MacDonald, C. Fyfe","doi":"10.1109/NNSP.2002.1030044","DOIUrl":"https://doi.org/10.1109/NNSP.2002.1030044","url":null,"abstract":"We investigate an extension of Hebbian learning in a principal component analysis network which has been derived to be optimal for a specific probability density function(PDF). We note that this probability density function is one of a family of PDFs and investigate the learning rules formed in order to be optimal for several members of this family. We show that, whereas previous authors have viewed the single member of the family as an extension of PCA, it is more appropriate to view the whole family of learning rules as methods of performing exploratory projection pursuit (EPP). We explore the performance of our method first in response to an artificial data type, then to a real data set.","PeriodicalId":117945,"journal":{"name":"Proceedings of the 12th IEEE Workshop on Neural Networks for Signal Processing","volume":"25 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-11-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122604475","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 : 2002-11-07DOI: 10.1109/NNSP.2002.1030100
N. Gilardi, Samy Bengio, M. Kanevski
This paper proposes the use of Gaussian mixture models to estimate conditional probability density functions in an environmental risk mapping context. A conditional Gaussian mixture model has been compared to, the geostatistical method of sequential Gaussian simulations and shows good performance in reconstructing the local PDF. The data sets used for this comparison are parts of the digital elevation model of Switzerland.
{"title":"Conditional Gaussian mixture models for environmental risk mapping","authors":"N. Gilardi, Samy Bengio, M. Kanevski","doi":"10.1109/NNSP.2002.1030100","DOIUrl":"https://doi.org/10.1109/NNSP.2002.1030100","url":null,"abstract":"This paper proposes the use of Gaussian mixture models to estimate conditional probability density functions in an environmental risk mapping context. A conditional Gaussian mixture model has been compared to, the geostatistical method of sequential Gaussian simulations and shows good performance in reconstructing the local PDF. The data sets used for this comparison are parts of the digital elevation model of Switzerland.","PeriodicalId":117945,"journal":{"name":"Proceedings of the 12th IEEE Workshop on Neural Networks for Signal Processing","volume":"48 11 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-11-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128568210","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 : 2002-11-07DOI: 10.1109/NNSP.2002.1030099
T. Schieder, C. Wilks, T. Rontzek, R. Eckmiller
We present a novel concept of a tactile communication system with dialog-based tuning possibilities for the exploration of tactile language developments. An experimental implementation of the proposed tactile intelligent sensory substitution system (TIS/sup 3/) is being tested in a closed loop set up with human subjects. TIS/sup 3/ consists of a tactile encoder (TE) to map desired objects onto a parallel stream of tactile stimulation time courses, a tactile stimulator (TS) to elicit spatio-temporal tactile sensations on a selected skin region, and a learning module (LM) to generate an ever improving parameter vector for TE based on the rating input of the human subject. As a first step, TE function was implemented as a set of 3/spl times/5 generators, which could be employed as a function of time by means of a DSP-based tunable spatio-temporal signal algorithm (TSA). Results showed that TIS/sup 3/ in a closed loop with humans subjects could be tuned to yield clearly distinguishable tactile perceptions within less than 80 iteration cycles.
{"title":"Towards a tunable tactile communication system: concept and first experiments","authors":"T. Schieder, C. Wilks, T. Rontzek, R. Eckmiller","doi":"10.1109/NNSP.2002.1030099","DOIUrl":"https://doi.org/10.1109/NNSP.2002.1030099","url":null,"abstract":"We present a novel concept of a tactile communication system with dialog-based tuning possibilities for the exploration of tactile language developments. An experimental implementation of the proposed tactile intelligent sensory substitution system (TIS/sup 3/) is being tested in a closed loop set up with human subjects. TIS/sup 3/ consists of a tactile encoder (TE) to map desired objects onto a parallel stream of tactile stimulation time courses, a tactile stimulator (TS) to elicit spatio-temporal tactile sensations on a selected skin region, and a learning module (LM) to generate an ever improving parameter vector for TE based on the rating input of the human subject. As a first step, TE function was implemented as a set of 3/spl times/5 generators, which could be employed as a function of time by means of a DSP-based tunable spatio-temporal signal algorithm (TSA). Results showed that TIS/sup 3/ in a closed loop with humans subjects could be tuned to yield clearly distinguishable tactile perceptions within less than 80 iteration cycles.","PeriodicalId":117945,"journal":{"name":"Proceedings of the 12th IEEE Workshop on Neural Networks for Signal Processing","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-11-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129224990","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 : 2002-11-07DOI: 10.1109/NNSP.2002.1030039
W. C. Siaw, S. L. Goh, A. I. Hanna, Christos Boukis, D. Mandic
A class of algorithms for training neural adaptive filters employed for nonlinear adaptive filtering is introduced. Sign algorithms incorporated with the fully adaptive normalised nonlinear gradient descent (SFANNGD) algorithm, normalised nonlinear gradient descent (SNNGD) algorithm and nonlinear gradient descent (SNGD) algorithm are proposed. The SFANNGD, SNNGD and the SNGD are derived based upon the principle of the sign algorithm used in the least mean square (LMS) filters. Experiments on nonlinear signals confirm that SFANNGD, SNNGD and the SNGD algorithms perform on par as compared to their basic algorithms but the sign algorithm decreases the overall computational complexity of the adaptive filter algorithms.
{"title":"Fully adaptive neural nonlinear FIR filters","authors":"W. C. Siaw, S. L. Goh, A. I. Hanna, Christos Boukis, D. Mandic","doi":"10.1109/NNSP.2002.1030039","DOIUrl":"https://doi.org/10.1109/NNSP.2002.1030039","url":null,"abstract":"A class of algorithms for training neural adaptive filters employed for nonlinear adaptive filtering is introduced. Sign algorithms incorporated with the fully adaptive normalised nonlinear gradient descent (SFANNGD) algorithm, normalised nonlinear gradient descent (SNNGD) algorithm and nonlinear gradient descent (SNGD) algorithm are proposed. The SFANNGD, SNNGD and the SNGD are derived based upon the principle of the sign algorithm used in the least mean square (LMS) filters. Experiments on nonlinear signals confirm that SFANNGD, SNNGD and the SNGD algorithms perform on par as compared to their basic algorithms but the sign algorithm decreases the overall computational complexity of the adaptive filter algorithms.","PeriodicalId":117945,"journal":{"name":"Proceedings of the 12th IEEE Workshop on Neural Networks for Signal Processing","volume":"35 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-11-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133112927","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 : 2002-09-04DOI: 10.1109/NNSP.2002.1030031
Deniz Erdoğmuş, J. Príncipe, K. Hild
Hebbian learning is one of the mainstays of biologically inspired neural processing. Hebb's (1949) rule is biologically plausible, and it has been extensively utilized in both computational neuroscience and in unsupervised training of neural systems. In these fields, Hebbian learning became synonymous for correlation learning. But it is known that correlation is a second order statistic of the data, so it is sub-optimal when the goal is to extract as much information as possible from the sensory data stream. We demonstrate how information learning can be implemented using Hebb's rule. Thus the paper brings a new understanding to how neural systems could, through Hebb's rule, extract information theoretic quantities rather than merely correlation.
{"title":"Do Hebbian synapses estimate entropy?","authors":"Deniz Erdoğmuş, J. Príncipe, K. Hild","doi":"10.1109/NNSP.2002.1030031","DOIUrl":"https://doi.org/10.1109/NNSP.2002.1030031","url":null,"abstract":"Hebbian learning is one of the mainstays of biologically inspired neural processing. Hebb's (1949) rule is biologically plausible, and it has been extensively utilized in both computational neuroscience and in unsupervised training of neural systems. In these fields, Hebbian learning became synonymous for correlation learning. But it is known that correlation is a second order statistic of the data, so it is sub-optimal when the goal is to extract as much information as possible from the sensory data stream. We demonstrate how information learning can be implemented using Hebb's rule. Thus the paper brings a new understanding to how neural systems could, through Hebb's rule, extract information theoretic quantities rather than merely correlation.","PeriodicalId":117945,"journal":{"name":"Proceedings of the 12th IEEE Workshop on Neural Networks for Signal Processing","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133994195","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 : 2002-02-11DOI: 10.1109/NNSP.2002.1030067
P. Hoyer
Non-negative sparse coding is a method for decomposing multivariate data into non-negative sparse components. We briefly describe the motivation behind this type of data representation and its relation to standard sparse coding and non-negative matrix factorization. We then give a simple yet efficient multiplicative algorithm for finding the optimal values of the hidden components. In addition, we show how the basis vectors can be learned from the observed data. Simulations demonstrate the effectiveness of the proposed method.
{"title":"Non-negative sparse coding","authors":"P. Hoyer","doi":"10.1109/NNSP.2002.1030067","DOIUrl":"https://doi.org/10.1109/NNSP.2002.1030067","url":null,"abstract":"Non-negative sparse coding is a method for decomposing multivariate data into non-negative sparse components. We briefly describe the motivation behind this type of data representation and its relation to standard sparse coding and non-negative matrix factorization. We then give a simple yet efficient multiplicative algorithm for finding the optimal values of the hidden components. In addition, we show how the basis vectors can be learned from the observed data. Simulations demonstrate the effectiveness of the proposed method.","PeriodicalId":117945,"journal":{"name":"Proceedings of the 12th IEEE Workshop on Neural Networks for Signal Processing","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-02-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127788939","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 : 1900-01-01DOI: 10.1109/NNSP.2002.1030028
M. V. Van Hulle
A new learning algorithm for kernel-based topographic map formation is introduced. The kernels are Gaussians, and their centers and ranges individually adapted so as to yield an equiprobabilistic topographic map. The converged map also generates a heteroscedastic Gaussian mixture model of the input density. This is verified for both synthetic and real-world examples, and compared with other algorithms for kernel-based topographic map formation.
{"title":"Kernel-based topographic map formation achieved with normalized Gaussian competition","authors":"M. V. Van Hulle","doi":"10.1109/NNSP.2002.1030028","DOIUrl":"https://doi.org/10.1109/NNSP.2002.1030028","url":null,"abstract":"A new learning algorithm for kernel-based topographic map formation is introduced. The kernels are Gaussians, and their centers and ranges individually adapted so as to yield an equiprobabilistic topographic map. The converged map also generates a heteroscedastic Gaussian mixture model of the input density. This is verified for both synthetic and real-world examples, and compared with other algorithms for kernel-based topographic map formation.","PeriodicalId":117945,"journal":{"name":"Proceedings of the 12th IEEE Workshop on Neural Networks for Signal Processing","volume":"43 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114648639","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：