Presentation 2007-12-14
Robust Estimation of Inverse Covariance Matrix by Shrinkage Technique
Masakazu IWAMURA, Shinichiro OMACHI, Hirotomo ASO,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In pattern recognition, important measures including quadratic discriminant function (QDF) and Mahalanobis distance (MD) depend on the inverse of a covariance matrix. Since estimation accuracy of the matrix roughly determines pattern recognition performance, we have proposed an estimating method using "a shrinkage technique." In this paper, by introducing sphericity test and block diagonalization to the method, we achieve simplification and performance improvement.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) small sample problem / eigen decomposition / bias of eigenvalue / sphericity test / shrinkage of dimensionality
Paper # PRMU2007-156
Date of Issue

Conference Information
Committee PRMU
Conference Date 2007/12/6(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) Robust Estimation of Inverse Covariance Matrix by Shrinkage Technique
Sub Title (in English)
Keyword(1) small sample problem
Keyword(2) eigen decomposition
Keyword(3) bias of eigenvalue
Keyword(4) sphericity test
Keyword(5) shrinkage of dimensionality
1st Author's Name Masakazu IWAMURA
1st Author's Affiliation Graduate School of Engineering, Osaka Prefecture University()
2nd Author's Name Shinichiro OMACHI
2nd Author's Affiliation Graduate School of Engineering, Tohoku University
3rd Author's Name Hirotomo ASO
3rd Author's Affiliation Graduate School of Engineering, Tohoku University
Date 2007-12-14
Paper # PRMU2007-156
Volume (vol) vol.107
Number (no) 384
Page pp.pp.-
#Pages 5
Date of Issue