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

Information Theoretic Interpretation and Design of Generalized Bayesian Learning Methods

Research Project

Project/Area Number 19K11825
Research Category

Grant-in-Aid for Scientific Research (C)

Allocation TypeMulti-year Fund
Section一般
Review Section Basic Section 60010:Theory of informatics-related
Research InstitutionToyohashi University of Technology

Principal Investigator

Watanabe Kazuho  豊橋技術科学大学, 工学(系)研究科(研究院), 准教授 (10506744)

Project Period (FY) 2019-04-01 – 2024-03-31
Project Status Completed (Fiscal Year 2023)
Budget Amount *help
¥4,290,000 (Direct Cost: ¥3,300,000、Indirect Cost: ¥990,000)
Fiscal Year 2023: ¥650,000 (Direct Cost: ¥500,000、Indirect Cost: ¥150,000)
Fiscal Year 2022: ¥650,000 (Direct Cost: ¥500,000、Indirect Cost: ¥150,000)
Fiscal Year 2021: ¥1,170,000 (Direct Cost: ¥900,000、Indirect Cost: ¥270,000)
Fiscal Year 2020: ¥910,000 (Direct Cost: ¥700,000、Indirect Cost: ¥210,000)
Fiscal Year 2019: ¥910,000 (Direct Cost: ¥700,000、Indirect Cost: ¥210,000)
Keywordsレート歪み関数 / 板倉・斎藤距離 / オンライン予測 / 局所変分近似 / 不感応パラメータ / 正則化パラメータ / L1トレンドフィルタ / 最適再構成分布 / ミニマックス予測 / ε不感応損失 / スパース推定 / 情報量規準 / 頑健性 / 変動二値情報源 / 一般化事後分布 / 経験ベイズ法 / 潜在変数モデル
Outline of Research at the Start

本研究では統計的機械学習法の性能を解析する学習理論研究や、その知見を応用した学習法の設計論の構築を行う。特に、ベイズ推測に基づく学習法であるベイズ学習を、情報理論の一分野であり歪み有りデータ圧縮の限界を明かにするレート歪み理論により捉えることで、その性能や限界を明らかにする。そして、ベイズ学習やその一般化において重要な構成要素である事前分布および事後分布の設計に関し、レート歪み理論による性能限界の特徴づけおよび限界に近い性能を実現する設計法を構築する。

Outline of Final Research Achievements

To establish a fundamental theory for the information theoretically principled design of statistical learning methods, we studied online time-series prediction, the optimal reconstruction distribution achieving the rate-distortion function, the estimation of the insensitivity in loss functions and the estimation of the regularization parameter in sparse estimation methods. For respective subproblems, we extended a minimax optimal prediction method for real valued data to the prediction of the time-varying probabilities from binary inputs, characterized the optimal reconstruction distribution achieving the rate-distortion function for Itakura-Saito distortion measure, analyzed the generalization errors in learning of the insensitive parameter, and developed an efficient approximate estimation method for the regularization parameter of L1 trend filtering, which extracts piece-wise linear trend from time-series data.

Academic Significance and Societal Importance of the Research Achievements

歪み有りデータ圧縮の限界であるレート歪み関数を達成する再構成分布を特徴づける例を追加しており、既存の結果との対比に用いることができる。再構成分布の最適化に対応する、ベイズの定理に基づく学習法は一般に計算困難性を伴う。具体的な時系列解析問題において効率的な近似法を構築し、その性質が実験的に、または一部理論的に明らかにされた。

Report

(6 results)
  • 2023 Annual Research Report   Final Research Report ( PDF )
  • 2022 Research-status Report
  • 2021 Research-status Report
  • 2020 Research-status Report
  • 2019 Research-status Report
  • Research Products

    (29 results)

All 2024 2023 2022 2021 2020 2019 Other

All Journal Article (6 results) (of which Peer Reviewed: 6 results,  Open Access: 3 results) Presentation (18 results) (of which Int'l Joint Research: 7 results,  Invited: 1 results) Book (1 results) Remarks (4 results)

  • [Journal Article] Unbiased Estimating Equation and Latent Bias Under f-Separable Bregman Distortion Measures2024

    • Author(s)
      Masahiro Kobayashi, Kazuho Watanabe
    • Journal Title

      IEEE Transactions on Information Theory

      Volume: - Issue: 8 Pages: 5763-5781

    • DOI

      10.1109/tit.2024.3366538

    • Related Report
      2023 Annual Research Report
    • Peer Reviewed / Open Access
  • [Journal Article] Approximate Empirical Bayes Estimation of the Regularization Parameter in l1 Trend Filtering2022

    • Author(s)
      Omae Akiharu、Watanabe Kazuho
    • Journal Title

      Proc. of 2022 IEEE International Symposium on Information Theory (ISIT)

      Volume: 1 Pages: 462-467

    • DOI

      10.1109/isit50566.2022.9834623

    • Related Report
      2022 Research-status Report
    • Peer Reviewed
  • [Journal Article] Generalized Dirichlet-process-means for f-separable distortion measures2021

    • Author(s)
      Masahiro Kobayashi, Kazuho Watanabe
    • Journal Title

      Neurocomputing

      Volume: 458 Pages: 667-689

    • DOI

      10.1016/j.neucom.2020.03.123

    • Related Report
      2021 Research-status Report 2020 Research-status Report
    • Peer Reviewed / Open Access
  • [Journal Article] Unbiased Estimation Equation under f-Separable Bregman Distortion Measures2021

    • Author(s)
      Masahiro Kobayashi, Kazuho Watanabe
    • Journal Title

      Proc. of 2020 IEEE Information Theory Workshop

      Volume: - Pages: 311-315

    • DOI

      10.1109/itw46852.2021.9457678

    • Related Report
      2021 Research-status Report
    • Peer Reviewed
  • [Journal Article] Statistical Learning of the Insensitive Parameter in Support Vector Models2021

    • Author(s)
      Kazuho Watanabe
    • Journal Title

      Proc. of IEEE International Symposium on Information Theory

      Volume: - Pages: 2501-2506

    • DOI

      10.1109/isit45174.2021.9518182

    • Related Report
      2021 Research-status Report
    • Peer Reviewed
  • [Journal Article] Minimax Online Prediction of Varying Bernoulli Process under Variational Approximation2019

    • Author(s)
      Kenta Konagayoshi, Kazuho Watanabe
    • Journal Title

      Proc. of the 11th Asian Conference on Machine Learning (ACML 2019)

      Volume: 101 Pages: 141-156

    • Related Report
      2019 Research-status Report
    • Peer Reviewed / Open Access
  • [Presentation] 逆数ダイバージェンスにおける推定方程式の不偏性とその多次元拡張2023

    • Author(s)
      小林真佐大,渡辺一帆
    • Organizer
      第46回情報理論とその応用シンポジウム
    • Related Report
      2023 Annual Research Report
  • [Presentation] 経験分布とリサンプリングを用いたfダイバージェンスに基づく推定2023

    • Author(s)
      小林真佐大,渡辺一帆
    • Organizer
      情報論的学習理論ワークショップ
    • Related Report
      2023 Annual Research Report
  • [Presentation] γクロスエントロピーを用いたロバストVAEの提案2023

    • Author(s)
      梶大介,小林真佐大,渡辺一帆
    • Organizer
      情報論的学習理論ワークショップ
    • Related Report
      2023 Annual Research Report
  • [Presentation] 逆ガウスモデルにおける推定方程式の不偏性とその一般化2023

    • Author(s)
      小林真佐大,渡辺一帆
    • Organizer
      電気・電子・情報関係学会東海支部連合大会
    • Related Report
      2023 Annual Research Report
  • [Presentation] 不感応パラメータ推定サポートベクトル回帰における最適な次数選択2023

    • Author(s)
      望月翔太,渡辺一帆
    • Organizer
      電気・電子・情報関係学会東海支部連合大会
    • Related Report
      2023 Annual Research Report
  • [Presentation] Rate-distortion theoretical views of Bayesian learning coefficients2023

    • Author(s)
      Kazuho Watanabe
    • Organizer
      IMSI Workshop: Bayesian Statistics and Statistical Learning
    • Related Report
      2023 Annual Research Report
  • [Presentation] Approximate Empirical Bayes Estimation of the Regularization Parameter in l1 Trend Filtering2022

    • Author(s)
      Akiharu Omae, Kazuho Watanabe
    • Organizer
      IEEE International Symposium on Information Theory
    • Related Report
      2021 Research-status Report
    • Int'l Joint Research
  • [Presentation] 変分近似による L1 トレンドフィルタリングの超パラメータ推定2021

    • Author(s)
      大前昭晴, 渡辺一帆
    • Organizer
      第44回情報理論とその応用シンポジウム
    • Related Report
      2021 Research-status Report
  • [Presentation] Unbiased Estimation Equation under f-Separable Bregman Distortion Measures2021

    • Author(s)
      Masahiro Kobayashi, Kazuho Watanabe
    • Organizer
      IEEE Information Theory Workshop (ITW2020)
    • Related Report
      2020 Research-status Report
    • Int'l Joint Research
  • [Presentation] Statistical Learning of the Insensitive Parameter in Support Vector Models2021

    • Author(s)
      Kazuho Watanabe
    • Organizer
      IEEE International Symposium on Information Theory (ISIT2021)
    • Related Report
      2020 Research-status Report
    • Int'l Joint Research
  • [Presentation] Discrete Optimal Reconstruction Distributions for Itakura-Saito Distortion Measure2020

    • Author(s)
      Kazuho Watanabe
    • Organizer
      IEEE International Symposium on Information Theory (ISIT2020)
    • Related Report
      2020 Research-status Report
    • Int'l Joint Research
  • [Presentation] Multi-Decoder RNN Autoencoder Based on Variational Bayes Method2020

    • Author(s)
      Daisuke Kaji, Kazuho Watanabe, Masahiro Kobayashi
    • Organizer
      IEEE International Joint Conference on Neural Networks (IJCNN2020)
    • Related Report
      2020 Research-status Report
    • Int'l Joint Research
  • [Presentation] Rate-distortion theoretic interpretation of Bayesian learning coefficients2020

    • Author(s)
      Kazuho Watanabe
    • Organizer
      Information Theory and Applications Workshop
    • Related Report
      2019 Research-status Report
    • Int'l Joint Research / Invited
  • [Presentation] f分離可能Bregman歪み尺度における推定方程式の不偏性と推定量の性質2019

    • Author(s)
      小林真佐大, 渡辺一帆
    • Organizer
      電子情報通信学会情報理論研究会
    • Related Report
      2019 Research-status Report
  • [Presentation] Unbiased Estimation Equation Under f-separable Extension of Squared and Itakura-Saito Distances2019

    • Author(s)
      Masahiro Kobayashi, Kazuho Watanabe
    • Organizer
      Data Science, Statistics and Visualization (DSSV)
    • Related Report
      2019 Research-status Report
    • Int'l Joint Research
  • [Presentation] ε不感応損失関数におけるεパラメータの推定量とその性質2019

    • Author(s)
      沖本卓也, 渡辺一帆
    • Organizer
      電気・電子・情報関係学会東海支部連合大会
    • Related Report
      2019 Research-status Report
  • [Presentation] Discreteness of Optimal Reconstruction Distributions in Lossy Compression under Itakura-Saito Distortion2019

    • Author(s)
      渡辺一帆
    • Organizer
      第42回情報理論とその応用シンポジウム
    • Related Report
      2019 Research-status Report
  • [Presentation] f分離可能ブレグマン歪み尺度に基づくロバストな非負値行列分解2019

    • Author(s)
      小林真佐大, 渡辺一帆
    • Organizer
      第42回情報理論とその応用シンポジウム
    • Related Report
      2019 Research-status Report
  • [Book] Variational Bayesian learning theory2019

    • Author(s)
      Shinichi Nakajima, Kazuho Watanabe, Masashi Sugiyama
    • Total Pages
      558
    • Publisher
      Cambridge University Press
    • Related Report
      2019 Research-status Report
  • [Remarks] 豊橋技術科学大学情報・知能工学系学習推論システム研究室Webサイト

    • URL

      http://www.lisl.cs.tut.ac.jp/

    • Related Report
      2023 Annual Research Report
  • [Remarks] 豊橋技術科学大学情報・知能工学系学習推論システム研究室

    • URL

      http://www.lisl.cs.tut.ac.jp/

    • Related Report
      2022 Research-status Report 2021 Research-status Report
  • [Remarks] 豊橋技術科学大学 情報・知能工学系 学習推論システム研究室

    • URL

      http://www.lisl.cs.tut.ac.jp/

    • Related Report
      2020 Research-status Report
  • [Remarks] 豊橋技術科学大学学習推論システム研究室ホームページ

    • URL

      http://www.lisl.cs.tut.ac.jp/

    • Related Report
      2019 Research-status Report

URL: 

Published: 2019-04-18   Modified: 2025-01-30  

Information User Guide FAQ News Terms of Use Attribution of KAKENHI

Powered by NII kakenhi