Presentation | 1998/7/24 On the Orthonormal Discriminant Vector Method in Small Training Sample Size Situations Takanobu Miyamoto, Yoshihiko Hamamoto, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | In the small-sample, high-dimensional setting, the estimation error of a covariance matrix causes practical difficulties in designing a feature extractor based on the Fisher criterion. In this paper, we try to reduce the estimation error by using the Toeplitz method or regularization method. Experimental results support our approach. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Feature extraction / orthonormal discriminant vector method / within-class scatter matrix / Toeplitz method / regularization |
Paper # | IE98-39,PRMU98-62,MVE98-62 |
Date of Issue |
Conference Information | |
Committee | PRMU |
---|---|
Conference Date | 1998/7/24(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) | On the Orthonormal Discriminant Vector Method in Small Training Sample Size Situations |
Sub Title (in English) | |
Keyword(1) | Feature extraction |
Keyword(2) | orthonormal discriminant vector method |
Keyword(3) | within-class scatter matrix |
Keyword(4) | Toeplitz method |
Keyword(5) | regularization |
1st Author's Name | Takanobu Miyamoto |
1st Author's Affiliation | Faculty of Engineering, Yamaguchi University() |
2nd Author's Name | Yoshihiko Hamamoto |
2nd Author's Affiliation | Faculty of Engineering, Yamaguchi University |
Date | 1998/7/24 |
Paper # | IE98-39,PRMU98-62,MVE98-62 |
Volume (vol) | vol.98 |
Number (no) | 206 |
Page | pp.pp.- |
#Pages | 8 |
Date of Issue |