Presentation 2007-11-19
Midpoint-Validation Method for Support Vector Machine
Shota IKUBO, Hiroki TAMURA, Koichi TANNO,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In recent years, support vector machine (abbr. SVM) is one of the most influential and powerful tools for solving classification. The most attractive notion of SVM is the idea of the large margin. However, many experiment results showed that the boundary line created by SVM has deviation. Therefore, SVM uses the cross-validation technique in many cases. In this paper, we propose the method of decreasing the deviation of SVM by creating a midpoint data. The proposed method creates midpoint data and adjusts parameter of SVM by midpoint data. We compare its performance with those of the original SVM, Multilayer Perceptron (abbr. MLP), Radial Basis Function Neural Network (abbr. RBF), and tested our proposed method on several 2-class pattern classification problems.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Support Vector Machine / Neural Network / Pattern Classification Problem
Paper # NC2007-65
Date of Issue

Conference Information
Committee NC
Conference Date 2007/11/11(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 Neurocomputing (NC)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Midpoint-Validation Method for Support Vector Machine
Sub Title (in English)
Keyword(1) Support Vector Machine
Keyword(2) Neural Network
Keyword(3) Pattern Classification Problem
1st Author's Name Shota IKUBO
1st Author's Affiliation Faculty of Engineering, University of MIYAZAKI()
2nd Author's Name Hiroki TAMURA
2nd Author's Affiliation Faculty of Engineering, University of MIYAZAKI
3rd Author's Name Koichi TANNO
3rd Author's Affiliation Faculty of Engineering, University of MIYAZAKI
Date 2007-11-19
Paper # NC2007-65
Volume (vol) vol.107
Number (no) 328
Page pp.pp.-
#Pages 4
Date of Issue