Presentation | 1998/6/19 A Feature Extraction Method Based on the Latent Variable Models Naonori UEDA, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | The classification performance of a recognition system depends mainly on feature extraction. Conventionally, pattern features have been extracted in some ad hoc manner based on a designer's intuition. It is quite important to develop more general and analytic feature extraction methods applicable for more complex recognition tasks. This report introduces a new feature extraction method based on the latent variable models. Although this method is similar to previous analytic ones like KL transformation and PCA in the sense of the dimensionality reduction, it is essentially different from these in that it probabilistically performs the dimensionality reduction. Since it is formulated as a probability model, it can compute the posterior probability for unknown data. Moreover, it can be naturally extended to a mixture model within the maxlimum likelihood framework and therefore local dimensionality reduction can be realized. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | latent variable models / feature extraction / EM algorithm / mixture models |
Paper # | PRMU98-47 |
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 Feature Extraction Method Based on the Latent Variable Models |
Sub Title (in English) | |
Keyword(1) | latent variable models |
Keyword(2) | feature extraction |
Keyword(3) | EM algorithm |
Keyword(4) | mixture models |
1st Author's Name | Naonori UEDA |
1st Author's Affiliation | NTT Communication Science Laboratories() |
Date | 1998/6/19 |
Paper # | PRMU98-47 |
Volume (vol) | vol.98 |
Number (no) | 127 |
Page | pp.pp.- |
#Pages | 8 |
Date of Issue |