Presentation 2013-11-13
Safe Screening Rule for Incremental Learning
Yoshiki SUZUKI, Shota OKUMURA, Kohei OGAWA, Ichiro TAKEUCHI,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Efficient optimization algorithm is required in online learning or other incremental learning scenario since the model must be repeatedly updated each time new data is available. In this paper, we propose an efficient algorithm for incremental sparse model learning with main focus on support vector machine (SVM). Since the difference between the two models before and after an incremental operation should be reasonably small, non-SVs in the former model tend to be non-SVs also in the latter model. Actually, many existing algorithm exploit this conjecture and use a heuristic for removing a portion of non-SVs in their incremental updating operation. In this paper, we introduce a screening algorithm that can safely remove a subset of non-SVs that are guaranteed to remained to be non-SVs after the incremental updating operation. We illustrate the effectiveness of our approach through numerical experiments.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Support Vector Machine / Screening Rule / Incremental Learning
Paper # IBISML2013-64
Date of Issue

Conference Information
Committee IBISML
Conference Date 2013/11/5(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) Safe Screening Rule for Incremental Learning
Sub Title (in English)
Keyword(1) Support Vector Machine
Keyword(2) Screening Rule
Keyword(3) Incremental Learning
1st Author's Name Yoshiki SUZUKI
1st Author's Affiliation Department of Engineering, Nagoya Institute of Technology()
2nd Author's Name Shota OKUMURA
2nd Author's Affiliation Department of Engineering, Nagoya Institute of Technology
3rd Author's Name Kohei OGAWA
3rd Author's Affiliation Department of Engineering, Nagoya Institute of Technology
4th Author's Name Ichiro TAKEUCHI
4th Author's Affiliation Department of Engineering, Nagoya Institute of Technology
Date 2013-11-13
Paper # IBISML2013-64
Volume (vol) vol.113
Number (no) 286
Page pp.pp.-
#Pages 6
Date of Issue