Presentation 2011-03-28
Enumerating Feature-Sets with Submodularity
Y. KAWAHARA, K. TSUDA, T. WASHIO, A. TAKEDA, S. MINATO,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Selecting relevant features is a fundamental task in machine learning. Although many approaches have been investigated so far, regularized-learning with sparsity-inducing norms, such as LASSO, would be one of the most promising ones. In this paper, we investigate a challenging problem beyond feature selection - feature-set enumeration, where we try to enumerate all ε-optimal feature-sets in the l_0-regularized feature selection with sub-modular measures. We develop a novel algorithm for this problem based on the branch-and-bound framework, where bounding and cutting are performed using the structure of binary decision diagrams (BDDs) and the submodularity of selection measures. The performance of the proposed algorithm is investigated through experiments with artificial datasets.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) feature selection / submodular maximization / binary decision diagram (BDD) / l_0-regularization
Paper # IBISML2010-113
Date of Issue

Conference Information
Committee IBISML
Conference Date 2011/3/21(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 Information-Based Induction Sciences and Machine Learning (IBISML)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Enumerating Feature-Sets with Submodularity
Sub Title (in English)
Keyword(1) feature selection
Keyword(2) submodular maximization
Keyword(3) binary decision diagram (BDD)
Keyword(4) l_0-regularization
1st Author's Name Y. KAWAHARA
1st Author's Affiliation The Institute of Scientific and Industrial Research (ISIR), Osaka University()
2nd Author's Name K. TSUDA
2nd Author's Affiliation National Institute of Advanced Industrial Science and Technology (AIST)
3rd Author's Name T. WASHIO
3rd Author's Affiliation The Institute of Scientific and Industrial Research (ISIR), Osaka University
4th Author's Name A. TAKEDA
4th Author's Affiliation Department of Administration Engineering, Keio University
5th Author's Name S. MINATO
5th Author's Affiliation Division of Computer Science, Hokkaido University
Date 2011-03-28
Paper # IBISML2010-113
Volume (vol) vol.110
Number (no) 476
Page pp.pp.-
#Pages 6
Date of Issue