• Search Research Projects
  • Search Researchers
  • How to Use
  1. Back to previous page

Development, sophistication and applications of new ensemble learning algorithms based on information theoretic view point.

Research Project

Project/Area Number 19700264
Research Category

Grant-in-Aid for Young Scientists (B)

Allocation TypeSingle-year Grants
Research Field Statistical science
Research InstitutionNara Institute of Science and Technology

Principal Investigator

TAKENOUCHI Takashi  Nara Institute of Science and Technology, 情報科学研究科, 助教授 (50403340)

Project Period (FY) 2007 – 2008
Project Status Completed (Fiscal Year 2008)
Budget Amount *help
¥2,890,000 (Direct Cost: ¥2,500,000、Indirect Cost: ¥390,000)
Fiscal Year 2008: ¥1,690,000 (Direct Cost: ¥1,300,000、Indirect Cost: ¥390,000)
Fiscal Year 2007: ¥1,200,000 (Direct Cost: ¥1,200,000)
Keywords分類 / パターン認識 / ブースティング / 集団学習 / 多値判別 / 誤り訂正符号 / 局所化尤度 / ロバスト
Research Abstract

本研究では近年注目を集めている2種類のアンサンブル学習,ブースティングとECOCに基づく多値判別に関する研究を行った.
(1)ブースティングに関しては統計的考察に基づく最もロバストなロス関数の考案とミスラベルを考慮した多値判別アルゴリズムを提案し,アルゴリズムの収束性などに関する考察を行った.パターン識別問題においては例題に外れ値などが含まれることが多く,このような状況に対してロバストなアルゴリズムは有効に働く.
(2)ECOCに基づく多値判別問題に関しては,確率モデルを用いて複数の2値判別器を組み合わせて多値判別器(デコーダー)を構成するための枠組みを提案した.提案した枠組みにおいては各2値判別器同士の相関関係を考慮してデコーダーを構成する事ができる.また提案手法では確率モデルをベースとして推定を行っており,対象とするデータに対して適応的にデコーダーを構成することができる.確率モデルをベースとしたことで,従来の手法に用いられている様々な拡張や正則化を容易に適用出来ることが提案手法の利点であり,提案手法は従来の最新の多値判別手法を優越する性能を示す事がわかった.

Report

(3 results)
  • 2008 Annual Research Report   Final Research Report ( PDF )
  • 2007 Annual Research Report
  • Research Products

    (41 results)

All 2009 2008 2007 Other

All Journal Article (17 results) (of which Peer Reviewed: 17 results) Presentation (24 results)

  • [Journal Article] A multi-class classification method based on decoding of binary classifiers2009

    • Author(s)
      T. Takenouchi, and S. ishii.
    • Journal Title

      Neural Computation, in press 21:7(未決定)

    • Related Report
      2008 Annual Research Report
    • Peer Reviewed
  • [Journal Article] Collaborative prediction by multiple Bayesian networks and its application to printer usage modeling.2008

    • Author(s)
      J. Hirayama, M. Nakatomi, T. Takenouchi and S. Ishii
    • Journal Title

      Behaviormetrika 35(2)

      Pages: 99-114

    • NAID

      110006950237

    • Related Report
      2008 Final Research Report
    • Peer Reviewed
  • [Journal Article] α-Bayesian collaboration of multiple predictors and its applications to hybrid recommendation and user modeling.2008

    • Author(s)
      J. Hirayama, M. Nakatomi, T. Takenouchi and S. Ishii
    • Journal Title

      Neural Information Processing Letters and Reviews 12(1-3)

      Pages: 11-20

    • Related Report
      2008 Final Research Report
    • Peer Reviewed
  • [Journal Article] Robust boosting algorithm against mislabeling in multi-class problems.2008

    • Author(s)
      T. Takenouchi, S. Eguchi, N. Murata, and T. Kanamori
    • Journal Title

      Neural Computation 20(16)

      Pages: 1596-1630

    • Related Report
      2008 Final Research Report
    • Peer Reviewed
  • [Journal Article] A probabilistic modeling of MOSAIC learning2008

    • Author(s)
      S. Osaga, J. Hirayama, T. Takenouchi and S. Ishii
    • Journal Title

      Artificial Life and Robotics 12(1-2)

      Pages: 167-171

    • Related Report
      2008 Final Research Report
    • Peer Reviewed
  • [Journal Article] Combining multiple decisions : applications to bioinformatics.2008

    • Author(s)
      N. Yukinawa, T. Takenouchi, S. Oba and S. Ishii
    • Journal Title

      Journal of Physics : Confenrece Series 95 012018

      Pages: 13-13

    • Related Report
      2008 Final Research Report
    • Peer Reviewed
  • [Journal Article] Collaborative prediction by multiple Bayesian networks and its applicati on to printer usage modeling2008

    • Author(s)
      Hirayama, J
    • Journal Title

      Behaviormetrika 35(2)

      Pages: 99-114

    • Related Report
      2008 Annual Research Report
    • Peer Reviewed
  • [Journal Article] α-Bayesian collaboration of multiple predictors and its applications to hybrid recommendation and user modeling2008

    • Author(s)
      Hirayama, J
    • Journal Title

      Neural Information Processing-Letters and Reviews 12(1-3)

      Pages: 11-20

    • Related Report
      2008 Annual Research Report
    • Peer Reviewed
  • [Journal Article] Robust boosting algorithm against mislabeling in multi-class problems2008

    • Author(s)
      T. Takenouchi
    • Journal Title

      To appear in Neural Computaion 20(16)

      Pages: 1596-1630

    • Related Report
      2007 Annual Research Report
    • Peer Reviewed
  • [Journal Article] Aprobabilistic modeling of MOSAIC learning<2008

    • Author(s)
      S. Osaga
    • Journal Title

      Artificial Life and Robotics 12(1-2)

      Pages: 167-171

    • Related Report
      2007 Annual Research Report
    • Peer Reviewed
  • [Journal Article] Combining multiple decisions: applications to bioinformatics2008

    • Author(s)
      N. Yukinawa
    • Journal Title

      Journal of Physics: Conference Series 95 012018

      Pages: 1-13

    • Related Report
      2007 Annual Research Report
    • Peer Reviewed
  • [Journal Article] Robust Loss Functions for Boosting.2007

    • Author(s)
      Kanamori, T., Takenouchi, T., Eguchi, S., and Murata, N.
    • Journal Title

      Neural Computation 19

      Pages: 2183-2244

    • Related Report
      2008 Final Research Report
    • Peer Reviewed
  • [Journal Article] Parameter Estimation for von Mises-Fisher distributions.2007

    • Author(s)
      Tanabe, A., Fukumizu, K., Oba, S., Takenouchi, T., and Ishii, S.
    • Journal Title

      Computational Statistics 22(1)

      Pages: 145-157

    • Related Report
      2008 Final Research Report
    • Peer Reviewed
  • [Journal Article] Group AdaBoost: accurate prediction and selection of importantgenes2007

    • Author(s)
      T. Takenouchi
    • Journal Title

      IPSJ Transactions on Bioinformatics 48,No.SIG5

      Pages: 1-8

    • Related Report
      2007 Annual Research Report
    • Peer Reviewed
  • [Journal Article] Robust Loss Functions for Boosting2007

    • Author(s)
      T. Kanamori
    • Journal Title

      Neural Computation 19

      Pages: 2183-2244

    • Related Report
      2007 Annual Research Report
    • Peer Reviewed
  • [Journal Article] Parameter Estimation for von Mises-Fisher Distributions2007

    • Author(s)
      A. Tanabe
    • Journal Title

      Computational Statistics 22(1)

      Pages: 145-157

    • Related Report
      2007 Annual Research Report
    • Peer Reviewed
  • [Journal Article] A multi-class classification method based on decoding of binary classifiers.

    • Author(s)
      T. Takenouchi
    • Journal Title

      Neural Comutation (in press)

    • Related Report
      2008 Final Research Report
    • Peer Reviewed
  • [Presentation] ECOCに基づく多値判別法とその拡張2009

    • Author(s)
      竹之内高志
    • Organizer
      大阪市立大学数学研究所 情報幾何研究集会2009 (OCAMI meeting on "Information Geometry 2009")
    • Place of Presentation
      大阪市立大
    • Year and Date
      2009-01-25
    • Related Report
      2008 Final Research Report
  • [Presentation] ECOCに基づく多値判別法とその拡張2009

    • Author(s)
      竹之内高志
    • Organizer
      大阪市立大学数学研究所情報幾何研究集会2009(OCAMI meeting on "Information Geometry 2009")
    • Place of Presentation
      大阪市立大
    • Year and Date
      2009-01-25
    • Related Report
      2008 Annual Research Report
  • [Presentation] Ternary Bradley-Terry model-based decoding for multi-class classification.2008

    • Author(s)
      T. Takenouchi
    • Organizer
      IEEE International Workshop on Machine Learning For Signal Processing
    • Place of Presentation
      Cancun, Mexico
    • Year and Date
      2008-10-18
    • Related Report
      2008 Final Research Report
  • [Presentation] Ternary Bradley-Terry model-based decoding for multi-class classification2008

    • Author(s)
      T. Takenouchi
    • Organizer
      IEEE International Workshop on Machine Learning For Signal Processing
    • Place of Presentation
      Cancun, Mexico
    • Year and Date
      2008-10-18
    • Related Report
      2008 Annual Research Report
  • [Presentation] ブースティングによるパターン識別2008

    • Author(s)
      竹之内高志
    • Organizer
      日本鉄鋼協会秋季講演大会
    • Place of Presentation
      熊本大学
    • Year and Date
      2008-09-24
    • Related Report
      2008 Annual Research Report 2008 Final Research Report
  • [Presentation] ダイナミクスのベイズ推定に基づくマルチモジュール型適応制御モデル2008

    • Author(s)
      大佐賀智
    • Organizer
      日本神経回路学会第18回全国大会
    • Place of Presentation
      産業技術総合研究所
    • Year and Date
      2008-09-24
    • Related Report
      2008 Annual Research Report 2008 Final Research Report
  • [Presentation] ECOC復号法に基づく階層的多値判別法2008

    • Author(s)
      大山賀己
    • Organizer
      ニューロコンピューティング研究会(NC)
    • Place of Presentation
      玉川大学
    • Year and Date
      2008-03-13
    • Related Report
      2008 Final Research Report 2007 Annual Research Report
  • [Presentation] An unified approach to collaborative and feature-based recommendation based on probabilistic latent semantic models.2008

    • Author(s)
      A. Ishihara
    • Organizer
      International Symposiumon Artificial Life and Robotics (AROB 13th '08)
    • Place of Presentation
      B-Con Plaza, Beppu, Oita, Japan
    • Year and Date
      2008-01-31
    • Related Report
      2007 Annual Research Report
  • [Presentation] An unified approach to collaborative and feature-based recommendation based on probabilistic latent semantic models.2008

    • Author(s)
      A. Ishihara
    • Organizer
      International Symposium on Artificial Life and Robotics (AROB 13th '08)
    • Place of Presentation
      B-Con Plaza, Beppu, Oita, Japan
    • Year and Date
      2008-01-30
    • Related Report
      2008 Final Research Report
  • [Presentation] アスペクトモデルに基づいた特徴ベース協調フィルタフィング2008

    • Author(s)
      石原愛美
    • Organizer
      人工知能学会研究会資料
    • Place of Presentation
      北海道大学
    • Year and Date
      2008-01-17
    • Related Report
      2008 Final Research Report
  • [Presentation] 1分子蛍光イメージングに対する尤度比検定を用いたアプローチ2008

    • Author(s)
      山下哲矢
    • Organizer
      ニューロコンピューティング研究会(NC)
    • Place of Presentation
      玉川大学
    • Year and Date
      2008-01-15
    • Related Report
      2008 Final Research Report
  • [Presentation] A multi-class classification with a probabilistic localized decoder.2007

    • Author(s)
      T. Takenouchi
    • Organizer
      IEEE International Symposium on Signal Processing and Information Technology (ISSPIT)
    • Place of Presentation
      Cairo, Egypt
    • Year and Date
      2007-12-17
    • Related Report
      2008 Final Research Report
  • [Presentation] A multi-class classification with a probabilistic localized decoder2007

    • Author(s)
      T. Takenouchi
    • Organizer
      IEEE International Symposium on Signal Processing and Information Technology
    • Place of Presentation
      Cairo, Egypt
    • Year and Date
      2007-12-17
    • Related Report
      2007 Annual Research Report
  • [Presentation] Bayesian collaborative predictors for general user modeling tasks.2007

    • Author(s)
      Hirayama, J.
    • Organizer
      International Conference on Neural Information Processing (ICONIP)
    • Place of Presentation
      Kita-Kyuushu, Japan
    • Year and Date
      2007-11-16
    • Related Report
      2008 Final Research Report
  • [Presentation] Bayesian collaborative predictors for general user modeling tasks2007

    • Author(s)
      J. Hirayama
    • Organizer
      International Conference on Neural Information Processing (ICONIP)
    • Place of Presentation
      Kitakyushu,Japan
    • Year and Date
      2007-11-16
    • Related Report
      2007 Annual Research Report
  • [Presentation] Combining multiple decisions : applications to bioinformatics.2007

    • Author(s)
      S. Ishii
    • Organizer
      The International Workshop on Statistical-Mechanical Informatics
    • Place of Presentation
      Paruru Plaza, Kyoto, Japan
    • Year and Date
      2007-09-18
    • Related Report
      2008 Final Research Report
  • [Presentation] Combining multiple decisions: applications to bioinformatics2007

    • Author(s)
      S. Ishii
    • Organizer
      The International Workshop on Statistical-Mechanical Informatics
    • Place of Presentation
      Paruru Plaza,Kyoto,Japan
    • Year and Date
      2007-09-18
    • Related Report
      2007 Annual Research Report
  • [Presentation] ECOCに基づく多値判別法の局所化2007

    • Author(s)
      竹之内高志
    • Organizer
      2007年度統計関連学会連合大会
    • Place of Presentation
      神戸大学
    • Year and Date
      2007-09-08
    • Related Report
      2008 Final Research Report 2007 Annual Research Report
  • [Presentation] A probabilistic decoding approach to multi-class classification.2007

    • Author(s)
      T. Takenouchi
    • Organizer
      Interantional Joint Conference on Neural Networks (IJCNN)
    • Place of Presentation
      Orlando, Florida, USA.
    • Year and Date
      2007-08-14
    • Related Report
      2008 Final Research Report
  • [Presentation] A probabilistic decoding approach to multi-class classiffcation2007

    • Author(s)
      T. Takenouchi
    • Organizer
      Interantional Joint Conference on Neural Networks(IJCNN)
    • Place of Presentation
      Orlando, Florida
    • Year and Date
      2007-08-14
    • Related Report
      2007 Annual Research Report
  • [Presentation] Multiclass classification as a decoding problem.2007

    • Author(s)
      T. Takenouchi
    • Organizer
      The First IEEE Symposium on Foundations of Computational Intelligence (FOCI'07)
    • Place of Presentation
      Hawaii, USA.
    • Year and Date
      2007-04-04
    • Related Report
      2008 Final Research Report
  • [Presentation] Multiclass classification as a decoding problem.2007

    • Author(s)
      T. Takenouchi
    • Organizer
      The First IEEE Symposium on Foundationsof Computational Intelligence (FOCI'07),
    • Place of Presentation
      Hawaii, USA
    • Year and Date
      2007-04-04
    • Related Report
      2007 Annual Research Report
  • [Presentation] A probabilistic model of MOSAIC.2007

    • Author(s)
      S. Osaga
    • Organizer
      The First IEEE Symposium on Foundations of Computational Intelligence (FOCI'07)
    • Place of Presentation
      Hawaii, USA.
    • Year and Date
      2007-04-02
    • Related Report
      2008 Final Research Report
  • [Presentation] A probabilistic model of MOSAIC.2007

    • Author(s)
      S. Osaga
    • Organizer
      The First IEEE Symposium on Foundations of Computational Intelligence (FOCI'07),
    • Place of Presentation
      Hawaii, USA
    • Year and Date
      2007-04-02
    • Related Report
      2007 Annual Research Report

URL: 

Published: 2007-04-01   Modified: 2016-04-21  

Information User Guide FAQ News Terms of Use Attribution of KAKENHI

Powered by NII kakenhi