Presentation 2006-11-23
Deterministic Network Learning Based on Classification-Error Rate
Kar-Ann Toh,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) This paper presents a new error counting network for pattern classification. Two major issues namely, the tedious iterative network learning issue and the ill-posed classification error counting issue have been addressed. Essentially, a linear series functional approximation to network structure and a smooth quadratic error counting function were proposed to resolve these two major computational issues. Our analysis shows that the new error counting objective can be related to the least-squares error objective by adjusting the class-specific normalization factors. The binary classification network is subsequently extended to cater for multi-category problems. An extensive empirical evaluation based on 42 UCI data sets validates the usefulness of the proposed method.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Pattern Classification / Neural Networks / Discriminant Functions / Polynomials and Machine Learning
Paper # PRMU2006-124
Date of Issue

Conference Information
Committee PRMU
Conference Date 2006/11/16(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 ENG
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Deterministic Network Learning Based on Classification-Error Rate
Sub Title (in English)
Keyword(1) Pattern Classification
Keyword(2) Neural Networks
Keyword(3) Discriminant Functions
Keyword(4) Polynomials and Machine Learning
1st Author's Name Kar-Ann Toh
1st Author's Affiliation Biometrics Engineering Research Center School of Electrical & Electronic Engineering Yonsei University()
Date 2006-11-23
Paper # PRMU2006-124
Volume (vol) vol.106
Number (no) 375
Page pp.pp.-
#Pages 10
Date of Issue