Presentation 2002/9/13
Feature Extraction Based on Support Vector Machine Decision Boundaries
Kotaro SHIMA, Masaru TODORIKI, Atsuyuki SUZUKI,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Traditionally, Principal Component Analysis method has been used to extract features from high dimensional data. However, this method does not make use of class information, thus does not necessarily extract effective features specific to classification task. In this paper, we propose a feature extraction method based on decision boundaries obtained by Support Vector Machines, which is gaining much attention in pattern recognition community. We demonstrate the effectiveness of this method using a benchmark dataset.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Feature Extraction / Support Vector Machine / Decision Boundary / Singular Value Decomposition
Paper # WIT2002-31
Date of Issue

Conference Information
Committee WIT
Conference Date 2002/9/13(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 Well-being Information Technology(WIT)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Feature Extraction Based on Support Vector Machine Decision Boundaries
Sub Title (in English)
Keyword(1) Feature Extraction
Keyword(2) Support Vector Machine
Keyword(3) Decision Boundary
Keyword(4) Singular Value Decomposition
1st Author's Name Kotaro SHIMA
1st Author's Affiliation Department of Quantum Engineering and Systems Science, University of Tokyo()
2nd Author's Name Masaru TODORIKI
2nd Author's Affiliation Department of Quantum Engineering and Systems Science, University of Tokyo
3rd Author's Name Atsuyuki SUZUKI
3rd Author's Affiliation Department of Quantum Engineering and Systems Science, University of Tokyo
Date 2002/9/13
Paper # WIT2002-31
Volume (vol) vol.102
Number (no) 320
Page pp.pp.-
#Pages 5
Date of Issue