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

2004 Fiscal Year Final Research Report Summary

Dimensionality Reduction for Designing Online Algorithms

Research Project

Project/Area Number 15500001
Research Category

Grant-in-Aid for Scientific Research (C)

Allocation TypeSingle-year Grants
Section一般
Research Field Fundamental theory of informatics
Research InstitutionTohoku University

Principal Investigator

TAKIMOTO Eiji  Tohoku University, Graduate School of Information Sciences, Associate Professor, 大学院・情報科学研究科, 助教授 (50236395)

Project Period (FY) 2003 – 2004
Keywordsonline prediction / kernel method / data compression / dimensionality reduction / Boosting / risk information
Research Abstract

A number of methods have been developed for predicting nearly as well as the best predictor among a set of experts. These methods have the same mechanism of making predictions that are based on the weighted average of experts' advices. In many natural applications, however, we have to deal with exponentially or infinitely many experts to be combined, and so it is computationally infeasible to explicitly maintain weights for all experts. In this research, we proposed a method of maintaining some parameter vector in a low dimensional space that implicitly represents the weight vector, with which we can efficiently simulate the weighted average prediction. Below are the major results obtained in this research project.
For the class of exponentially many predictors associated with the paths of a graph, we gave a method of efficiently simulating the weighted average prediction by maintaining probabilistic flows on the edges. This gives a new kernel called the path kernel which turned out to be useful in many applications.
We proposed a new scheme of Boosting by dividing and merging the domain repeatedly to form a decision diagram as its final hypothesis. This gives a unified framework in which we can now analyze the AdaBoost-type and the Decision Tree-type algorithms that were thought to be derived from quite different principles.
We generalized the model so that the learner is allowed to see the bounds on the losses of experts (risk information) and gave a tight performance bound of the Aggregating Algorithm.

  • Research Products

    (28 results)

All 2005 2004 2003

All Journal Article (28 results)

  • [Journal Article] リスク情報を用いたオンライン資源分配2005

    • Author(s)
      原田薫明
    • Journal Title

      信学会技報・コンピュテーション 2004-86

      Pages: 105-110

    • Description
      「研究成果報告書概要(和文)」より
  • [Journal Article] 【招待】オンライン予測の理論と応用2005

    • Author(s)
      瀧本英二
    • Journal Title

      第18回回路とシステム軽井沢ワークショップ 18

      Pages: 595-600

    • Description
      「研究成果報告書概要(和文)」より
  • [Journal Article] Online Allocation with Risk Information2005

    • Author(s)
      Shigeaki Harada
    • Journal Title

      Proc.4th Japanese-Hungarian Symposium On Discrete Mathematics And Its Applications (掲載予定)

    • Description
      「研究成果報告書概要(和文)」より
  • [Journal Article] On Proper Learning for Monotone Term Decision Lists from Queries2005

    • Author(s)
      Eiji Takimoto
    • Journal Title

      Proc.Workshop on Learning with Logic and Logic for Learning (掲載予定)

    • Description
      「研究成果報告書概要(和文)」より
  • [Journal Article] Online Allocation with Risk Information2005

    • Author(s)
      Shigeaki Harada
    • Journal Title

      Technical Report of IEICE COMP2004-86

      Pages: 105-110

    • Description
      「研究成果報告書概要(欧文)」より
  • [Journal Article] Online prediction -Theory and Applications2005

    • Author(s)
      Eiji Takimoto
    • Journal Title

      Proc.18th Workshop on Circuits and Systems in Karuizawa

      Pages: 595-600

    • Description
      「研究成果報告書概要(欧文)」より
  • [Journal Article] Online Allocation with Risk Information2005

    • Author(s)
      Shigeaki Harada
    • Journal Title

      Proc.4th Japanese-Hungarian Symp.on Discrete Math.and Its Applications (To appear in)

    • Description
      「研究成果報告書概要(欧文)」より
  • [Journal Article] On Proper Learning for Monotone Term Decision Lists from Queries2005

    • Author(s)
      Eiji Takimoto
    • Journal Title

      Proc.Workshop on Learning with Logic and Logic for Learning (To appear in)

    • Description
      「研究成果報告書概要(欧文)」より
  • [Journal Article] ブール関数のPTF表現の複雑さについて2004

    • Author(s)
      松尾健史
    • Journal Title

      信学会技報・コンピュテーション 2003-87

      Pages: 9-16

    • Description
      「研究成果報告書概要(和文)」より
  • [Journal Article] Boosting Based on Divide and Merge2004

    • Author(s)
      Eiji Takimoto
    • Journal Title

      Lecture Notes in Artificial Intelligence 3244

      Pages: 127-141

    • Description
      「研究成果報告書概要(和文)」より
  • [Journal Article] On the PTF expression of Boolean functions2004

    • Author(s)
      Kenshi Matsuo
    • Journal Title

      Technical Report of IEICE COMP2003-87

      Pages: 9-16

    • Description
      「研究成果報告書概要(欧文)」より
  • [Journal Article] Boosting based on Divide and Merge2004

    • Author(s)
      Eiji Takimoto
    • Journal Title

      LNAI 3244

      Pages: 127-141

    • Description
      「研究成果報告書概要(欧文)」より
  • [Journal Article] Path Kernels and Multiplicative Updates2003

    • Author(s)
      Eiji Takimoto
    • Journal Title

      Journal of Machine Learning Research 4

      Pages: 773-818

    • Description
      「研究成果報告書概要(和文)」より
  • [Journal Article] Relationships between Horn formulas and XOR-MDNF formulas2003

    • Author(s)
      Kenshi Matsuo
    • Journal Title

      IEICE Transactions on Information and Systems E87-D(2)

      Pages: 343-351

    • Description
      「研究成果報告書概要(和文)」より
  • [Journal Article] Top-down decision tree learning as information based boosting2003

    • Author(s)
      Eiji Takimoto
    • Journal Title

      Theoretical Computer Science 292・2

      Pages: 447-464

    • Description
      「研究成果報告書概要(和文)」より
  • [Journal Article] オンライン学習の学習曲線に関する研究2003

    • Author(s)
      原田薫明
    • Journal Title

      信学会技報・コンピュテーション 2003-15

      Pages: 45-51

    • Description
      「研究成果報告書概要(和文)」より
  • [Journal Article] ガウス分布推定問題に対するミニマックス戦略2003

    • Author(s)
      瀧本英二
    • Journal Title

      信学会技報・コンピュテーション 2003-41

      Pages: 29-36

    • Description
      「研究成果報告書概要(和文)」より
  • [Journal Article] 最終段ミニマックスアルゴリズム2003

    • Author(s)
      瀧本英二
    • Journal Title

      信学会技報・コンピュテーション 2003-59

      Pages: 39-44

    • Description
      「研究成果報告書概要(和文)」より
  • [Journal Article] 分割と併合に基づくブースティング2003

    • Author(s)
      竹内寛明
    • Journal Title

      情報科学技術フォーラム情報技術レターズ 2

      Pages: 19-20

    • Description
      「研究成果報告書概要(和文)」より
  • [Journal Article] 多値分類問題に対するブースティングの困難さについて2003

    • Author(s)
      田中恭
    • Journal Title

      情報科学技術フォーラム一般講演論文集 A-054

      Pages: 119-120

    • Description
      「研究成果報告書概要(和文)」より
  • [Journal Article] Path Kernels and Multiplicative Updates2003

    • Author(s)
      Eiji Takimoto
    • Journal Title

      Journal of Machine Learning 4

      Pages: 773-818

    • Description
      「研究成果報告書概要(欧文)」より
  • [Journal Article] Relationships between Horn formulas and XOR-MDNF formulas2003

    • Author(s)
      Kenshi Matsuo
    • Journal Title

      IEICE Trans.on Info.& Syst. E87-D(2)

      Pages: 343-351

    • Description
      「研究成果報告書概要(欧文)」より
  • [Journal Article] Top-down decision tree learning as information based Boosting2003

    • Author(s)
      Eiji Takimoto
    • Journal Title

      Theoretical Computer Science 292(2)

      Pages: 447-464

    • Description
      「研究成果報告書概要(欧文)」より
  • [Journal Article] Learning Curve Bounds for Online Learning2003

    • Author(s)
      Shigeaki Harada
    • Journal Title

      Technical Report of IEICE COMP2003-15

      Pages: 45-51

    • Description
      「研究成果報告書概要(欧文)」より
  • [Journal Article] The Minimax Strategy for Gaussian Density Estimation2003

    • Author(s)
      Eiji Takimoto
    • Journal Title

      Technical Report of IEICE COMP2003-41

      Pages: 29-36

    • Description
      「研究成果報告書概要(欧文)」より
  • [Journal Article] The Last-step Minimax Algorithm2003

    • Author(s)
      Eiji Takimoto
    • Journal Title

      Technical Report of IEICE COMP2003-59

      Pages: 39-44

    • Description
      「研究成果報告書概要(欧文)」より
  • [Journal Article] Boosting based on Divide and Merge2003

    • Author(s)
      Hiroaki Takeuchi
    • Journal Title

      Information Technology Letters 2

      Pages: 19-20

    • Description
      「研究成果報告書概要(欧文)」より
  • [Journal Article] On the Difficulty of Boosting For Multi-class Classification Problems2003

    • Author(s)
      Kyo Tanaka
    • Journal Title

      Proc.2nd Forum on Information Technology A-054

      Pages: 119-120

    • Description
      「研究成果報告書概要(欧文)」より

URL: 

Published: 2006-07-11  

Information User Guide FAQ News Terms of Use Attribution of KAKENHI

Powered by NII kakenhi