Presentation 2001/3/15
Minimum Classification Risk Criterion for the Competitive Learning : Optimization of the Error-Reject Tradeoff
Susumu KUBOTA, Hiroyuki MIZUTANI, Yoshiaki KUROSAWA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) A criterion for the competitive learning is proposed. Though MCE criterion has been commonly used, it doesn't lead to optimal classification performance nor is suitable for steepest descent method. Considering reject cost and introducing average classification risk, quantitative relation between loss function, which is minimized by learning, and classification performance is given. The proposed criterion not only leads to good classification performance but significantly reduce time and effort required for the learning.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) pattern recognition / competitive learning / reject / classification risk minimization
Paper # PRMU2000-208
Date of Issue

Conference Information
Committee PRMU
Conference Date 2001/3/15(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) Minimum Classification Risk Criterion for the Competitive Learning : Optimization of the Error-Reject Tradeoff
Sub Title (in English)
Keyword(1) pattern recognition
Keyword(2) competitive learning
Keyword(3) reject
Keyword(4) classification risk minimization
1st Author's Name Susumu KUBOTA
1st Author's Affiliation TOSHIBA Corporation R&D Center()
2nd Author's Name Hiroyuki MIZUTANI
2nd Author's Affiliation TOSHIBA Corporatio Digital Media Network Company
3rd Author's Name Yoshiaki KUROSAWA
3rd Author's Affiliation TOSHIBA Corporatio Digital Media Network Company
Date 2001/3/15
Paper # PRMU2000-208
Volume (vol) vol.100
Number (no) 701
Page pp.pp.-
#Pages 8
Date of Issue