Presentation 1996/3/18
S/N Analysis of Associative Memory Model Learned by Correlation Learning using Non-monotonic Property
Koji Koizumi, Masato Okada,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) We analyze the associative memory model learned by the correlation learnig using the non-monotonic unit. It has been known that the model has better properties than the conventional one. We propose the most simple model which can be considered to have property and then analyze the model by the so-called signal-to-noise analysis. The model has the twice larger storage capacity than the convetional one even in the case of the monotonic units are used in the recall process. Not only that the variance of the cross-talk noise is smaller, but also that the model has negative correlation between the signal term and the cross-talk noise. Therefore, the variance of the internal potential becomes much smaller than the cross-talk noise.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) associative memory model / non-monotonic unit / correlation learning / orthogonal-projection matrix / signal-to-noise analysis
Paper # NC-95-132
Date of Issue

Conference Information
Committee NC
Conference Date 1996/3/18(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Neurocomputing (NC)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) S/N Analysis of Associative Memory Model Learned by Correlation Learning using Non-monotonic Property
Sub Title (in English)
Keyword(1) associative memory model
Keyword(2) non-monotonic unit
Keyword(3) correlation learning
Keyword(4) orthogonal-projection matrix
Keyword(5) signal-to-noise analysis
1st Author's Name Koji Koizumi
1st Author's Affiliation Department of Biophysical Engineering, Faculty of Engineering Science, Osaka University()
2nd Author's Name Masato Okada
2nd Author's Affiliation Department of Biophysical Engineering, Faculty of Engineering Science, Osaka University
Date 1996/3/18
Paper # NC-95-132
Volume (vol) vol.95
Number (no) 598
Page pp.pp.-
#Pages 8
Date of Issue