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

グラフニューラルネットワークの理論解析と高速化

Research Project

Project/Area Number 22KJ1703
Project/Area Number (Other) 21J22490 (2021-2022)
Research Category

Grant-in-Aid for JSPS Fellows

Allocation TypeMulti-year Fund (2023)
Single-year Grants (2021-2022)
Section国内
Review Section Basic Section 61030:Intelligent informatics-related
Research InstitutionKyoto University

Principal Investigator

佐藤 竜馬  京都大学, 情報学研究科, 特別研究員(DC1)

Project Period (FY) 2023-03-08 – 2024-03-31
Project Status Completed (Fiscal Year 2023)
Budget Amount *help
¥2,200,000 (Direct Cost: ¥2,200,000)
Fiscal Year 2023: ¥700,000 (Direct Cost: ¥700,000)
Fiscal Year 2022: ¥700,000 (Direct Cost: ¥700,000)
Fiscal Year 2021: ¥800,000 (Direct Cost: ¥800,000)
Keywordsグラフニューラルネットワーク / 機械学習
Outline of Research at the Start

グラフデータは様々な構造を記述することができる。例えば、ソーシャルネットワークは人のつながりを記述し、知識グラフは概念のつながりを記述し、化合物グラフは原子の繋がりを記述できる。近年のグラフニューラルネットワークの急速な発展に伴い、これらの構造化されたデータに対する予測の性能が大いに向上した。しかし、グラフニューラルネットワークはヒューリスティックに利用されることが多く、原理の深い解明には至っていない。本研究課題では、グラフニューラルネットワークの原理を理論的に明らかにし、性能の向上および計算の高速化をめざす。

Outline of Annual Research Achievements

本年度は既存のグラフニューラルネットワークの常識を打ち破る新たな理論を打ち立てることに成功した研究 "Graph Neural Networks can Recover the Hidden Features Solely from the Graph Structure" が国際会議 International Conference on Machine Learning (ICML) に採択され、現地で発表を行った。グラフニューラルネットワークはノード特徴量を受け取り、これをグラフ上で滑らかにすることで性能が得られると考えられてきた。ノード特徴量が利用できない場合のグラフニューラルネットワークの性能は理論的に解明されてこなかった。本研究では、ノード特徴量が利用できない場合であっても、グラフニューラルネットワークはグラフ構造を表すノード特徴量を自力で生み出し、これを元に予測を行なうことができることを示した。グラフニューラルネットワークの理論解析という本研究課題に直接繋がる大きな成果といえる。こちらの研究内容についてはギリシャのグラフニューラルネットワークの研究チームからも依頼を受けて講演を行った。また、グラフニューラルネットワークについての教科書『グラフニューラルネットワーク (機械学習プロフェッショナルシリーズ)』 https://www.amazon.co.jp/dp/4065347823 を執筆した。こちらの教科書はグラフニューラルネットワークはどのような場面で役に立つのかなどといった基本的なことから、本課題と直接関係のあるグラフニューラルネットワークの理論解析と高速化についても個別に章を設けて詳細に解説してある。

Report

(3 results)
  • 2023 Annual Research Report
  • 2022 Annual Research Report
  • 2021 Annual Research Report
  • Research Products

    (25 results)

All 2024 2023 2022 2021

All Journal Article (4 results) (of which Int'l Joint Research: 1 results,  Peer Reviewed: 4 results,  Open Access: 3 results) Presentation (19 results) (of which Int'l Joint Research: 15 results,  Invited: 3 results) Book (2 results)

  • [Journal Article] Momentum Tracking: Momentum Acceleration for Decentralized Deep Learning on Heterogeneous Data2023

    • Author(s)
      Yuki Takezawa, Han Bao, Kenta Niwa, Ryoma Sato, Makoto Yamada
    • Journal Title

      Transactions on Machine Learning Research (TMLR)

      Volume: -

    • Related Report
      2023 Annual Research Report
    • Peer Reviewed / Open Access
  • [Journal Article] Approximating 1-Wasserstein Distance with Trees2022

    • Author(s)
      Makoto Yamada, Yuki Takezawa, Ryoma Sato, Han Bao, Zornitsa Kozareva, Sujith Ravi:
    • Journal Title

      Transactions on Machine Learning Research

      Volume: -

    • Related Report
      2022 Annual Research Report
    • Peer Reviewed / Open Access / Int'l Joint Research
  • [Journal Article] Constant Time Graph Neural Networks2022

    • Author(s)
      Ryoma Sato, Makoto Yamada, Hisashi Kashima
    • Journal Title

      ACM Transactions on Knowledge Discovery from Data (TKDD)

      Volume: 16 Issue: 5 Pages: 1-31

    • DOI

      10.1145/3502733

    • Related Report
      2022 Annual Research Report 2021 Annual Research Report
    • Peer Reviewed / Open Access
  • [Journal Article] Poincare: Recommending Publication Venues via Treatment Effect Estimation2022

    • Author(s)
      Ryoma Sato, Makoto Yamada, Hisashi Kashima
    • Journal Title

      Journal of Informetrics

      Volume: 16 Issue: 2 Pages: 101283-101283

    • DOI

      10.1016/j.joi.2022.101283

    • Related Report
      2022 Annual Research Report 2021 Annual Research Report
    • Peer Reviewed
  • [Presentation] Active Learning from the Web2023

    • Author(s)
      Ryoma Sato
    • Organizer
      The Web Conference (WWW)
    • Related Report
      2023 Annual Research Report 2022 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Graph Neural Networks can Recover the Hidden Features Solely from the Graph Structure2023

    • Author(s)
      Ryoma Sato
    • Organizer
      International Conference on Machine Learning (ICML)
    • Related Report
      2023 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Beyond Exponential Graph: Communication-Efficient Topologies for Decentralized Learning via Finite-time Convergence2023

    • Author(s)
      Yuki Takezawa*, Ryoma Sato*, Han Bao, Kenta Niwa, Makoto Yamada
    • Organizer
      Conference on Neural Information Processing Systems (NeurIPS)
    • Related Report
      2023 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Towards Principled User-side Recommender Systems2022

    • Author(s)
      Ryoma Sato
    • Organizer
      ACM International Conference on Information and Knowledge Management (CIKM)
    • Related Report
      2022 Annual Research Report
    • Int'l Joint Research
  • [Presentation] CLEAR: A Fully User-side Image Search System2022

    • Author(s)
      Ryoma Sato
    • Organizer
      ACM International Conference on Information and Knowledge Management (CIKM)
    • Related Report
      2022 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Twin Papers: A Simple Framework of Causal Inference for Citations via Coupling2022

    • Author(s)
      Ryoma Sato, Makoto Yamada, Hisashi Kashima
    • Organizer
      ACM International Conference on Information and Knowledge Management (CIKM)
    • Related Report
      2022 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Feature Robust Optimal Transport for High-dimensional Data2022

    • Author(s)
      Mathis Petrovich, Chao Liang, Ryoma Sato, Yanbin Liu, Yao-Hung Hubert Tsai, Linchao Zhu, Yi Yang, Ruslan Salakhutdinov, Makoto Yamada
    • Organizer
      European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML-PKDD)
    • Related Report
      2022 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Re-evaluating Word Mover’s Distance2022

    • Author(s)
      Ryoma Sato, Makoto Yamada, Hisashi Kashima
    • Organizer
      International Conference on Machine Learning (ICML)
    • Related Report
      2022 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Word Tour: One-dimensional Word Embeddings via the Traveling Salesman Problem2022

    • Author(s)
      Ryoma Sato
    • Organizer
      Annual Conference of the North American Chapter of the Association for Computational Linguistics (NAACL)
    • Related Report
      2022 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Private Recommender Systems: How Can Users Build Their Own Fair Recommender Systems without Log Data?2022

    • Author(s)
      Ryoma Sato
    • Organizer
      SIAM International Conference on Data Mining (SDM)
    • Related Report
      2022 Annual Research Report 2021 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Private Recommender Systems: How Can Users Build Their Own Fair Recommender Systems without Log Data?2022

    • Author(s)
      佐藤竜馬
    • Organizer
      日本ソフトウェア科学会
    • Related Report
      2022 Annual Research Report
    • Invited
  • [Presentation] Enumerating Fair Packages for Group Recommendations2022

    • Author(s)
      Ryoma Sato
    • Organizer
      International Conference on Web Search and Data Mining (WSDM)
    • Related Report
      2021 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Retrieving Black-box Optimal Images from External Databases2022

    • Author(s)
      Ryoma Sato
    • Organizer
      International Conference on Web Search and Data Mining (WSDM)
    • Related Report
      2021 Annual Research Report
    • Int'l Joint Research
  • [Presentation] ユーザーサイド情報検索システム2022

    • Author(s)
      佐藤竜馬
    • Organizer
      第6回統計・機械学習若手シンポジウム
    • Related Report
      2021 Annual Research Report
    • Invited
  • [Presentation] Fixed Support Tree-Sliced Wasserstein Barycenter2022

    • Author(s)
      Yuki Takezawa, Ryoma Sato, Zornitsa Kozareva, Sujith Ravi, Makoto Yamada
    • Organizer
      International Conference on Artificial Intelligence and Statistics (AISTATS)
    • Related Report
      2021 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Random Features Strengthen Graph Neural Networks2021

    • Author(s)
      Ryoma Sato, Makoto Yamada, Hisashi Kashima
    • Organizer
      SIAM International Conference on Data Mining (SDM)
    • Related Report
      2021 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Supervised Tree-Wasserstein Distance2021

    • Author(s)
      Yuki Takezawa, Ryoma Sato, Makoto Yamada
    • Organizer
      International Conference on Machine Learning (ICML)
    • Related Report
      2021 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Tree-Wasserstein距離のための教師あり学習2021

    • Author(s)
      竹澤祐貴, 佐藤竜馬, 山田誠.
    • Organizer
      第24回情報論的学習理論ワークショップ
    • Related Report
      2021 Annual Research Report
  • [Presentation] 最適輸送入門2021

    • Author(s)
      佐藤竜馬
    • Organizer
      第24回情報論的学習理論ワークショップ
    • Related Report
      2021 Annual Research Report
    • Invited
  • [Book] グラフニューラルネットワーク (機械学習プロフェッショナルシリーズ)2024

    • Author(s)
      佐藤竜馬
    • Total Pages
      336
    • Publisher
      講談社
    • ISBN
      4065347823
    • Related Report
      2023 Annual Research Report
  • [Book] 最適輸送の理論とアルゴリズム2023

    • Author(s)
      佐藤 竜馬
    • Total Pages
      320
    • Publisher
      講談社
    • ISBN
      4065305144
    • Related Report
      2022 Annual Research Report

URL: 

Published: 2021-05-27   Modified: 2024-12-25  

Information User Guide FAQ News Terms of Use Attribution of KAKENHI

Powered by NII kakenhi