Presentation | 2010-09-05 A Study on Feature Selection Path for High-Dimensional Local Classifiers Ichiro TAKEUCHI, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | We study feature selection and weighting problems for local-based classifier. The proposed algorithm is formulated as a regularized empirical loss minimization problem w.r.t. feature weights. We show that the solution of the problem yields sparse feature weights, and it enables us to select a subset of features by choosing an appropriate regularization parameter. Furthermore, we develop an algorithm to compute the entire regularization path of the problem. Feature weighting problem is defined using, what is called, target-neighbors, and they must be updated according to the change of feature weights. The proposed algorithm can detect target-neighbor changes by exploiting the piecewise-linearity of the regularization path. This leads us to construct a regularization path algorithm with target-neighbor updates. We show the effectiveness of the proposed algorithm through numerical experiments. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | local-based classifier / nearest-neighbor classifier / feature selection / feature weighting / convex optimization / path-following / parametric programming / metric learning |
Paper # | PRMU2010-71,IBISML2010-43 |
Date of Issue |
Conference Information | |
Committee | PRMU |
---|---|
Conference Date | 2010/8/29(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 Study on Feature Selection Path for High-Dimensional Local Classifiers |
Sub Title (in English) | |
Keyword(1) | local-based classifier |
Keyword(2) | nearest-neighbor classifier |
Keyword(3) | feature selection |
Keyword(4) | feature weighting |
Keyword(5) | convex optimization |
Keyword(6) | path-following |
Keyword(7) | parametric programming |
Keyword(8) | metric learning |
1st Author's Name | Ichiro TAKEUCHI |
1st Author's Affiliation | Department of Engineering, Nagoya Institute of Technology() |
Date | 2010-09-05 |
Paper # | PRMU2010-71,IBISML2010-43 |
Volume (vol) | vol.110 |
Number (no) | 187 |
Page | pp.pp.- |
#Pages | 8 |
Date of Issue |