Presentation 2001/10/11
A Multiclass Classification Method Based on Multiple Pairwise Classifiers
Tomoyuki HAMAMURA, Hiroyuki MIZUTANI, Bunpei IRIE,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In this paper, a multiclass classification method is presented. To classify multiple classes, we combine two-class classifiers of all pair of classes. To combine two-class classifiers, we propose a Resemblance Model to calculate a posteriori probability. We showed the validity of this model by using approximation of a posteriori probability formula. When we have a posteriori probability, we can apply the Bayes classification rule to obtain the optimal decision, and furthermore, control the rejection quite easy. As an example, we performed a experiment of handwritten numerals recognition and obtained the good result.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Pattern recognition / Two-class classification / Multiclass classification / Resemblance Model / Posteriori probability / Bayes classification rule
Paper # PRMU2001-102,NC2001-52
Date of Issue

Conference Information
Committee PRMU
Conference Date 2001/10/11(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) A Multiclass Classification Method Based on Multiple Pairwise Classifiers
Sub Title (in English)
Keyword(1) Pattern recognition
Keyword(2) Two-class classification
Keyword(3) Multiclass classification
Keyword(4) Resemblance Model
Keyword(5) Posteriori probability
Keyword(6) Bayes classification rule
1st Author's Name Tomoyuki HAMAMURA
1st Author's Affiliation TOSHIBA Corp.()
2nd Author's Name Hiroyuki MIZUTANI
2nd Author's Affiliation TOSHIBA Corp.
3rd Author's Name Bunpei IRIE
3rd Author's Affiliation TOSHIBA Corp.
Date 2001/10/11
Paper # PRMU2001-102,NC2001-52
Volume (vol) vol.101
Number (no) 362
Page pp.pp.-
#Pages 6
Date of Issue