講演抄録/キーワード |
講演名 |
2010-11-05 15:30
[ポスター講演]Regularization Strategies and Empirical Bayesian Learning for MKL ○Ryota Tomioka・Taiji Suzuki(Univ. of Tokyo) IBISML2010-100 |
抄録 |
(和) |
Multiple kernel learning (MKL) has received considerable attention recently. In this paper, we show how different MKL algorithms can be understood as applications of different types of regularization on the kernel weights. We show that many algorithms based on Ivanov regularization, have their corresponding Tikhonov regularization formulations. In addition, we show that the two regularization strategies are connected by the block-norm formulation. The Tikhonov-regularization-based formulation of MKL allows us to consider a generative probabilistic model behind MKL. Based on this model, we propose learning algorithms for the kernel weights through the maximization of marginalized likelihood. |
(英) |
Multiple kernel learning (MKL) has received considerable attention recently. In this paper, we show how different MKL algorithms can be understood as applications of different types of regularization on the kernel weights. We show that many algorithms based on Ivanov regularization, have their corresponding Tikhonov regularization formulations. In addition, we show that the two regularization strategies are connected by the block-norm formulation. The Tikhonov-regularization-based formulation of MKL allows us to consider a generative probabilistic model behind MKL. Based on this model, we propose learning algorithms for the kernel weights through the maximization of marginalized likelihood. |
キーワード |
(和) |
/ / / / / / / |
(英) |
Multiple Kernel Learning / MKL / Regularization / Empirical Bayesian Learning / Evidence / / / |
文献情報 |
信学技報, vol. 110, no. 265, IBISML2010-100, pp. 303-310, 2010年11月. |
資料番号 |
IBISML2010-100 |
発行日 |
2010-10-28 (IBISML) |
ISSN |
Print edition: ISSN 0913-5685 Online edition: ISSN 2432-6380 |
著作権に ついて |
技術研究報告に掲載された論文の著作権は電子情報通信学会に帰属します.(許諾番号:10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
PDFダウンロード |
IBISML2010-100 |