Presentation 2004/3/10
Evaluation of Self-Organizing Maps based on Multinomial Distribution
Noriaki FUKUDA, Kazumi SAITO, Masumi ISHIKAWA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In this paper, we study how similarity definitions and neighborhood definitions in self-organizing maps (SOM) affect their classification performance. Conventional SOM uses a combination of a standard Euclidean norm for document similarity and a neighborhood definition based on a Gaussian distribution. We propose a novel SOM which adopts logarithmic likelihood based on multinomial distributions for document similarity, and neighborhood definition based on multinomial distributions. In computer experiments using three kinds of benchmark document data sets, we evaluate these SOM models from the points of view of the classification performances and resulting self-organizing maps.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) self-organizing map / text mining / multinomial distribution
Paper # NC2003-143
Date of Issue

Conference Information
Committee NC
Conference Date 2004/3/10(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) Evaluation of Self-Organizing Maps based on Multinomial Distribution
Sub Title (in English)
Keyword(1) self-organizing map
Keyword(2) text mining
Keyword(3) multinomial distribution
1st Author's Name Noriaki FUKUDA
1st Author's Affiliation Nagaoka University of Technology()
2nd Author's Name Kazumi SAITO
2nd Author's Affiliation NTT Communication Science Laboratories
3rd Author's Name Masumi ISHIKAWA
3rd Author's Affiliation Kyushu Institute of Technology
Date 2004/3/10
Paper # NC2003-143
Volume (vol) vol.103
Number (no) 732
Page pp.pp.-
#Pages 6
Date of Issue