Machine Learning under Dynamically Changing Environment

2010 Fiscal Year Self-evaluation Report

• Project/Area Number: 20680007
• Research Category: Grant-in-Aid for Young Scientists (A)
• Allocation Type: Single-year Grants
• Research Field: Intelligent informatics
• Research Institution: Tokyo Institute of Technology
• Principal Investigator: SUGIYAMA Masashi Tokyo Institute of Technology, 大学院・情報理工学研究科, 准教授 (90334515)
• Project Period (FY): 2008 – 2011
• Keywords: 機械学習 / データマイニング / 非定常環境 / 重点サンプリング / 密度比推定

Research Abstract

従来のデータ学習法は,データが生成される環境が定常的(即ち時間とともに変化しない)という大前提の下で理論が構築され,その理論に基づいたアルゴリズムが開発されてきた.しかしながら,近年の応用分野では定常性が成り立たない場合が多い.本研究プロジェクトの目的は,環境の変化に対応するための基礎理論,および,実用的なアルゴリズム開発を行なう.そして,その成果をロボティックスやブレイン・コンピュータインターフェースなどに応用する.

Research Products

• [Journal Article] Application of covariate shift adaptation techniques in brain computer interfaces. (2010)
  • Author(s): Li, Y., Kambara, H., Koike, Y., Sugiyama, M.
  • Journal Title: IEEE Transactions on Biomedical Engineering vol.57, no.6 Pages: 1318-1324
  • URL: http://sugiyama-www.cs.titech.ac.jp/~sugi/2010/BIWLDA.pdf
• [Journal Article] Efficient exploration through active learning for value function approximation in reinforcement learning. (2010)
  • Author(s): Akiyama, T., Hachiya, H., Sugiyama, M.
  • Journal Title: NeuralNetworks vol.23, no.5 Pages: 639-648
  • URL: http://sugiyama-www.cs.titech.ac.jp/~sugi/2010/API.pdf
• [Journal Article] Semi-supervised speaker identification under covariate shift. (2010)
  • Author(s): Yamada, M., Sugiyama, M., Matsui, T.
  • Journal Title: Signal Processing vol.90, no.8 Pages: 2353-2361
  • URL: http://sugiyama-www.cs.titech.ac.jp/~sugi/2010/IWKLR.pdf
• [Journal Article] Adaptive importance sampling for value function approximation in off-policy reinforcement learning. (2009)
  • Author(s): Hachiya, H., Akiyama, T., Sugiyama, M., Peters, J.
  • Journal Title: Neural Networks vol.22, no.10 Pages: 1399-1410
  • URL: http://sugiyama-www.cs.titech.ac.jp/~sugi/2009/SRPI.pdf
• [Journal Article] Pool-based active learning in approximate linear regression. (2009)
  • Author(s): Sugiyama, M., Nakajima, S.
  • Journal Title: Machine Learning vol.75, no.3 Pages: 249-274
  • URL: http://sugiyama-www.cs.titech.ac.jp/~sugi/2009/PALICE.pdf
• [Journal Article] A least-squares approach to direct importance estimation. (2009)
  • Author(s): Kanamori, T., Hido, S., Sugiyama, M.
  • Journal Title: Journal of Machine Learning Research vol.10 (Jul.) Pages: 1391-1445
  • URL: http://sugiyama-www.cs.titech.ac.jp/~sugi/2009/LSIF.pdf
• [Journal Article] Direct importance estimation for covariate shift adaptation. (2008)
  • Author(s): Sugiyama, M., Suzuki, T., Nakajima, S., Kashima, H., von Bunau, P., Kawanabe, M.
  • Journal Title: Annals of the Institute of Statistical Mathematics vol.60, no.4 Pages: 699-746
  • URL: http://sugiyama-www.cs.titech.ac.jp/~sugi/2008/KLIEP.pdf
• [Presentation] Perceived age estimation under lighting condition change by covariate shift adaptation. (2010)
  • Author(s): Ueki, K., Sugiyama, M., Ihara, Y.
  • Organizer: 20th International Conferenceon Pattern Recognition (ICPR2010)
  • URL: http://sugiyama-www.cs.titech.ac.jp/~sugi/2010/ICPR2010a.pdf
  • Place of Presentation: Istanbul, Turkey
  • Year and Date: 20100823-20100826
• [Presentation] Efficient sample reuse in EM-based policy search. (2009)
  • Author(s): Hachiya, H., Peters, J., Sugiyama, M.
  • Organizer: European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases(ECML-PKDD2009)
  • URL: http://sugiyama-www.cs.titech.ac.jp/~sugi/2009/ECML-PKDD2009.pdf
  • Place of Presentation: Bled, Slovenia
  • Year and Date: 20090907-20090911
• [Presentation] Active policy iteration: Efficient exploration through active learning for value function approximation in reinforcement learning. (2009)
  • Author(s): Akiyama, T., Hachiya, H., Sugiyama, M.
  • Organizer: Twenty-FirstInternational Joint Conference onArtificial Intelligence (IJCAI2009)
  • URL: http://sugiyama-www.cs.titech.ac.jp/~sugi/2009/IJCAI2009.pdf
  • Place of Presentation: Pasadena, California
  • Year and Date: 20090711-20090717
• [Presentation] Efficient direct density ratio estimation for non-stationarity adaptation and outlier detection (2008)
  • Author(s): Kanamori, T., Hido, S., Sugiyama, M.
  • Organizer: Neural Information Processing Systems (NIPS2008)
  • URL: http://sugiyama-www.cs.titech.ac.jp/~sugi/2008/NIPS2008.pdf
  • Place of Presentation: Vancouver, Canada
  • Year and Date: 20081208-20081213
• [Book] MIT Press, Cambridge, MA, USA (2009)
  • Author(s): Quinonero-Candela, J., Sugiyama, M., Schwaighofer, A., Lawrence, N.D.(Eds.)
  • Publisher: Dataset Shift in Machine Learning
• [Remarks] ソフトウェアの公開ページ
  • URL: http://sugiyama-www.cs.titech.ac.jp/~sugi/software/index.html