Presentation 1994/11/29
The Relationship between the Probability Density of Input Signals and the Reference Vector Distribution in a Modified Kohonen's Model
Kouji Kurata,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) It is a property common to the models of self-organizing topographic mapping to assign more units to the signals which they learn more frequently than others in order to get higher resolution about such signals.In Kohonen's model,Ritter showed that resolution or the density of units mapped into the signal space is asymptotically proportional,not to the probability density function p of input signals,but to p^2, 3>.In this paper, a class of models will be proposed which contains Kohonen's model as a special case.Within the class,it is possible to design such a model that the density of units is asymptotically proportional to p^δ with any given positive δ.It will be proved that the mapping loses its stability if one tries to make δ negative.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Self-Organizing Map / Reference Vector / Instability / Micro- Structure
Paper # NC94-50
Date of Issue

Conference Information
Committee NC
Conference Date 1994/11/29(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 ENG
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) The Relationship between the Probability Density of Input Signals and the Reference Vector Distribution in a Modified Kohonen's Model
Sub Title (in English)
Keyword(1) Self-Organizing Map
Keyword(2) Reference Vector
Keyword(3) Instability
Keyword(4) Micro- Structure
1st Author's Name Kouji Kurata
1st Author's Affiliation Faculty of Engineering Science,Osaka University()
Date 1994/11/29
Paper # NC94-50
Volume (vol) vol.94
Number (no) 377
Page pp.pp.-
#Pages 8
Date of Issue