Presentation 1998/6/19
A New Effective Quadratic Discriminant Function on Multi Dimensions
Mitsuru SAKAI, Masaaki YONEDA, Hiroyuki HASE, Hiroshi MARUYAMA, Michiko NAOE,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) We propose a new quadratic discriminant function. Fisrt, we show that the eigenvalues of a covariance matrix obtained from samples are biased. Then, we describe how to rectify them and how to use the rectified eigenvalues and discuss why they work well. We derive simultaneous linear equations to calculate approximates of the rectified eigenvalues in a multi-dimensional normal case by using perturbation theory. Finally, we show by Monte Carlo method that our disctiminant function works effectively in eight-dimentional normal cases especially in the case of a small sample size.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) training sample size / covariance matrix / bias of eigenvalue / normal distribution / quadratic discriminant function / perturbation theory
Paper # PRMU98-43
Date of Issue

Conference Information
Committee PRMU
Conference Date 1998/6/19(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 New Effective Quadratic Discriminant Function on Multi Dimensions
Sub Title (in English)
Keyword(1) training sample size
Keyword(2) covariance matrix
Keyword(3) bias of eigenvalue
Keyword(4) normal distribution
Keyword(5) quadratic discriminant function
Keyword(6) perturbation theory
1st Author's Name Mitsuru SAKAI
1st Author's Affiliation Faculty of Engineering, Toyama University()
2nd Author's Name Masaaki YONEDA
2nd Author's Affiliation Faculty of Engineering, Toyama University
3rd Author's Name Hiroyuki HASE
3rd Author's Affiliation Faculty of Engineering, Toyama University
4th Author's Name Hiroshi MARUYAMA
4th Author's Affiliation Faculty of Engineering, Toyama University
5th Author's Name Michiko NAOE
5th Author's Affiliation Tateyama System Laboratory Co., Ltd.
Date 1998/6/19
Paper # PRMU98-43
Volume (vol) vol.98
Number (no) 127
Page pp.pp.-
#Pages 8
Date of Issue