Presentation 1998/6/19
Optimal Selection of the Number of Components in Classifiers based on Mixture Models
Hiroshi TENMOTO, Mineichi KUDO, Masaru SHIMBO,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) We propose a new method for selection of the number of components of a mixture model for pattern classifiers. In previous methods, the optimal number of components is selected independently for each class. When the main concern is classification and only a limited number of training samples is available, consideration of class-separability may be helpful. One selection criterion is based on a classification-oriented MDL(Minimum Description Length)principle. In this method, we evaluate each classifier by both the classification accuracy for the training samples and the complexity of the classifier itself. We select the optimal combination of the number of components by a trade-off between the two factors. The effectiveness of the proposed method was confirmed by experiments using artificial and real data.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Mixture Model / Classifier / Minimum Description Length Principle
Paper # PRMU98-42
Date of Issue

Conference Information
Committee PRMU
Conference Date 1998/6/19(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 Pattern Recognition and Media Understanding (PRMU)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Optimal Selection of the Number of Components in Classifiers based on Mixture Models
Sub Title (in English)
Keyword(1) Mixture Model
Keyword(2) Classifier
Keyword(3) Minimum Description Length Principle
1st Author's Name Hiroshi TENMOTO
1st Author's Affiliation Division of Systems and Information Engineering Graduate School of Engineering Hokkaido University()
2nd Author's Name Mineichi KUDO
2nd Author's Affiliation Division of Systems and Information Engineering Graduate School of Engineering Hokkaido University
3rd Author's Name Masaru SHIMBO
3rd Author's Affiliation Division of Systems and Information Engineering Graduate School of Engineering Hokkaido University
Date 1998/6/19
Paper # PRMU98-42
Volume (vol) vol.98
Number (no) 127
Page pp.pp.-
#Pages 5
Date of Issue