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

Estimation with extended models using generalized divergences and its applications

Research Project

Project/Area Number 16K00051
Research Category

Grant-in-Aid for Scientific Research (C)

Allocation TypeMulti-year Fund
Section一般
Research Field Statistical science
Research InstitutionFuture University-Hakodate

Principal Investigator

Takenouchi Takashi  公立はこだて未来大学, システム情報科学部, 准教授 (50403340)

Project Period (FY) 2016-04-01 – 2020-03-31
Project Status Completed (Fiscal Year 2019)
Budget Amount *help
¥4,420,000 (Direct Cost: ¥3,400,000、Indirect Cost: ¥1,020,000)
Fiscal Year 2018: ¥1,430,000 (Direct Cost: ¥1,100,000、Indirect Cost: ¥330,000)
Fiscal Year 2017: ¥1,430,000 (Direct Cost: ¥1,100,000、Indirect Cost: ¥330,000)
Fiscal Year 2016: ¥1,560,000 (Direct Cost: ¥1,200,000、Indirect Cost: ¥360,000)
Keywords一般化ダイバージェンス / 離散確率モデル / 非負値行列因子分解 / ロバスト / 非正規化モデル / 漸近有効性 / アンサンブル学習 / 多値判別 / 変形ブレグマン擬距離 / 情報幾何 / 統計的推定 / ダイバージェンス / 非確率モデル / 機械学習
Outline of Final Research Achievements

The estimation for probabilistic models in a discrete space requires the computation of the normalization term to be a probability. However, in high-dimensional discrete spaces, the computation of the normalization term often requires a computational complexity of exponential order. In this study, we use an un-normalized non-probabilistic model (called the extended model) instead of a probabilistic model, and proposed an estimator by combining the extended model with the technique of empirical localization and homogeneous divergence, which can be constructed without the normalization term and asymptotically achieves the Cramer-Rao bound.
We also construct a unified framework for handling multi-class classification methods and robust non-negative matrix factorization algorithm, by using generalized divergences.

Academic Significance and Societal Importance of the Research Achievements

提案した推定量は正規化項の計算が不要であるうえに,漸近的にクラメール・ラオの下限を達成するため, 通常の推定量と比較して数十-数百分の一の計算コストで最尤推定量に匹敵する性能を達成可能である.
また, 多値判別手法を扱うための統一的な枠組みは多くの従来手法を特殊ケースとして含むため, 性能に関する理論的な考察や比較が用意になった. 非負値行列因子分解法については, 再下降性と呼ばれる性質を手法に付与することが出来たため, ノイズに対して強力な頑健性をもたせることが可能となった.

Report

(5 results)
  • 2019 Annual Research Report   Final Research Report ( PDF )
  • 2018 Research-status Report
  • 2017 Research-status Report
  • 2016 Research-status Report
  • Research Products

    (17 results)

All 2019 2018 2017 2016

All Journal Article (4 results) (of which Peer Reviewed: 4 results,  Open Access: 1 results) Presentation (13 results) (of which Int'l Joint Research: 3 results,  Invited: 1 results)

  • [Journal Article] Statistical modeling of robust non-negative matrix factorization based on γ-divergence and its applications.2019

    • Author(s)
      K. Machida, and T. Takenouchi
    • Journal Title

      Japanese Journal of Statistics and Data Science

      Volume: 2 Issue: 2 Pages: 441-464

    • DOI

      10.1007/s42081-019-00041-3

    • Related Report
      2019 Annual Research Report
    • Peer Reviewed
  • [Journal Article] Binary classifiers ensemble based on Bregman divergence for multi-class classification2018

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

      Neurocomputing

      Volume: 273(17) Pages: 424-434

    • DOI

      10.1016/j.neucom.2017.08.004

    • Related Report
      2018 Research-status Report
    • Peer Reviewed
  • [Journal Article] Statistical Inference with Unnormalized Discrete Models and Localized Homogeneous Divergences.2017

    • Author(s)
      Takashi Takenouchi, and Takafumi Kanamori
    • Journal Title

      The Journal of Machine Learning Research

      Volume: 18 Pages: 1-26

    • Related Report
      2017 Research-status Report
    • Peer Reviewed / Open Access
  • [Journal Article] Graph-based Composite Local Bregman Divergences on Discrete Sample Spaces2017

    • Author(s)
      T. Kanamori, T. Takenouchi
    • Journal Title

      Neural Networks

      Volume: 95 Pages: 44-56

    • DOI

      10.1016/j.neunet.2017.06.005

    • Related Report
      2017 Research-status Report
    • Peer Reviewed
  • [Presentation] 非正規化モデルを用いた推定量とその性質2019

    • Author(s)
      竹之内高志
    • Organizer
      応用統計学会
    • Related Report
      2019 Annual Research Report
    • Invited
  • [Presentation] Causal Outcome Prediction on Combinatorial Action Spaces2019

    • Author(s)
      谷本 啓, 坂井 智哉, 竹之内 高志, 鹿島 久嗣
    • Organizer
      情報論的学習理論ワークショップ(IBIS2019)
    • Related Report
      2019 Annual Research Report
  • [Presentation] Robust contrastive learning and nonlinear ICA in the presence of outliers2019

    • Author(s)
      佐々木 博昭, 竹之内 高志, R. Monti, A. Hyvarinen
    • Organizer
      情報論的学習理論ワークショップ(IBIS2019)
    • Related Report
      2019 Annual Research Report
  • [Presentation] Zero-shot Domain Adaptation Based on Attribute Information2019

    • Author(s)
      M. Ishii, T. Takenouchi, and M. Sugiyama
    • Organizer
      The Eleventh Asian Conference on Machine Learning
    • Related Report
      2019 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Parameter Estimation with Generalized Empirical Localization.2019

    • Author(s)
      T. Takenouchi.
    • Organizer
      Geometric Science of Information
    • Related Report
      2019 Annual Research Report
    • Int'l Joint Research
  • [Presentation] 非正規化モデルを用いた推定法.2018

    • Author(s)
      竹之内高志
    • Organizer
      大規模複雑データの理論と方法総合的研究
    • Related Report
      2018 Research-status Report
  • [Presentation] Relaxing Imbalance with Positiveness2018

    • Author(s)
      谷本啓, 山田聡,竹之内高志,鹿島久嗣.
    • Organizer
      第 21 回情報論的学習理論ワークショップ(IBIS2018)
    • Related Report
      2018 Research-status Report
  • [Presentation] 異なるモデルの尤度関数の結合2018

    • Author(s)
      江口真透, 竹之内高志
    • Organizer
      統計関連学会連合大会
    • Related Report
      2018 Research-status Report
  • [Presentation] Parameter Estimation with Deformed Bregman Divergence2016

    • Author(s)
      竹之内高志
    • Organizer
      Information Geometry and its Applications IV
    • Place of Presentation
      Liblice, Czech Republic
    • Year and Date
      2016-06-13
    • Related Report
      2016 Research-status Report
    • Int'l Joint Research
  • [Presentation] 変形ブレグマンダイバージェンスを用いたパラメーター推定2016

    • Author(s)
      竹之内高志
    • Organizer
      2016年度 統計関連学会連合大会
    • Place of Presentation
      金沢大学角間キャンパス(石川県金沢市)
    • Related Report
      2016 Research-status Report
  • [Presentation] 離散空間上のグラフ構造に基づく局所ブレグマンダイバージェンス2016

    • Author(s)
      金森 敬文, 竹之内高志
    • Organizer
      2016年度 統計関連学会連合大会
    • Place of Presentation
      金沢大学角間キャンパス(石川県金沢市)
    • Related Report
      2016 Research-status Report
  • [Presentation] 変形ブレグマン擬距離とその応用2016

    • Author(s)
      竹之内高志
    • Organizer
      第19回情報論的学習理論ワークショップ
    • Place of Presentation
      京都大学(京都府京都市)
    • Related Report
      2016 Research-status Report
  • [Presentation] グラフ上の局所ブレグマンダイバージェンスによる統計的推論2016

    • Author(s)
      金森 敬文, 竹之内高志
    • Organizer
      第19回情報論的学習理論ワークショップ
    • Place of Presentation
      京都大学(京都府京都市)
    • Related Report
      2016 Research-status Report

URL: 

Published: 2016-04-21   Modified: 2021-02-19  

Information User Guide FAQ News Terms of Use Attribution of KAKENHI

Powered by NII kakenhi