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

2024 Fiscal Year Annual Research Report

Robust AI by Integration of Knowledge Representation and Machine Learning

Research Project

Project/Area Number 21H04905
Research InstitutionNational Institute of Informatics

Principal Investigator

井上 克巳  国立情報学研究所, 情報学プリンシプル研究系, 教授 (10252321)

Co-Investigator(Kenkyū-buntansha) 坂間 千秋  和歌山大学, システム工学部, 教授 (20273873)
Phua Yin・Jun  東京科学大学, 情報理工学院, 助教 (20963747)
Nicolas Schwind  国立研究開発法人産業技術総合研究所, 情報・人間工学領域, 主任研究員 (60646397)
小島 諒介  京都大学, 医学研究科, 講師 (70807651)
Project Period (FY) 2021-04-05 – 2025-03-31
Keywords知識表現 / 推論 / 機械学習 / 論理プログラミング / ロバストAI / ニューロシンボリックAI
Outline of Annual Research Achievements

本研究では、機械学習(ML)と知識表現・推論(KR)の両技術を有機的に統合することで、説明可能でありロバスト性も有するようなAI技術基盤の確立を目指し、(A)KR技術の導入によるML技術の説明可能性・更新容易性の向上、(B)ML技術に支えられたロバストなKR技術の開発、(C)MLとKRの統合による画期的なAI応用、という3つの研究目標を設定し研究を進めてきた。研究最終年の2024年度はこれらの3研究を統合し、KRとMLを統合したプラットフォームを提示することを目標とし、以下のような進捗があった。
・統合技術では、記号推論における不変性や対称性を用いて、より大規模な論理プログラムの学習が可能な深層学習モデルを提案した。また画像生成モデルを用いて未知なデータを生成し、継続学習をする画像認識モデルのロバスト性を向上させる技術を提案した。さらに、安定性に関する制約を付加した時系列深層モデルの開発を進め、画像処理における人流解析に制約モデルを導入した場合や、ブラックボックス最適化に制約を加えた場合など、派生的手法への応用も進めた。ニューロシンボリック学習では、深層学習等で画像等の正解ラベルではなく、画像を認識した後で推論を用いて得られる解をラベルとする問題に対し、ベクトル空間における微分可能な論理プログラム学習が有効であることを実証した。
・新たな応用では、議論をモデル化する際に、受け入れ可能な主張と論駁をメタ論理的に表現する新しい枠組を導入し、第3者的立場の視点や議論参加者の信念変化や自己欺瞞的な心理状態を表現可能とした。また、異なるMLモデルから学習する際の挙動を解釈するために、知識発展に対するモデルを開発することで学習の振る舞いを効率的に表現するための手法を開発した。またネットワーク内における通信において、各エージェントがどのように信念を更新すべきかについて新たな考察を加えた。

Research Progress Status

令和6年度が最終年度であるため、記入しない。

Strategy for Future Research Activity

令和6年度が最終年度であるため、記入しない。

  • Research Products

    (39 results)

All 2025 2024 Other

All Int'l Joint Research (3 results) Journal Article (19 results) (of which Int'l Joint Research: 7 results,  Peer Reviewed: 19 results,  Open Access: 12 results) Presentation (15 results) (of which Int'l Joint Research: 12 results,  Invited: 4 results) Remarks (2 results)

  • [Int'l Joint Research] LaBRI CNRS / University of Bordeaux/LS2N / Ecole Centrale de Nantes/CRIL CNRS / University of Artois(フランス)

    • Country Name
      FRANCE
    • Counterpart Institution
      LaBRI CNRS / University of Bordeaux/LS2N / Ecole Centrale de Nantes/CRIL CNRS / University of Artois
  • [Int'l Joint Research] University of Potsdam(ドイツ)

    • Country Name
      GERMANY
    • Counterpart Institution
      University of Potsdam
  • [Int'l Joint Research] University of Bergen(ノルウェー)

    • Country Name
      NORWAY
    • Counterpart Institution
      University of Bergen
  • [Journal Article] Learning possibilistic dynamic systems from state transitions2025

    • Author(s)
      Hu Hongbo, Wang Yisong, Inoue Katsumi
    • Journal Title

      Fuzzy Sets and Systems

      Volume: 504 Pages: 109259~109259

    • DOI

      10.1016/j.fss.2024.109259

    • Peer Reviewed / Int'l Joint Research
  • [Journal Article] Learning Deep Dissipative Dynamics2025

    • Author(s)
      Okamoto Yuji, Kojima Ryosuke
    • Journal Title

      Proceedings of the AAAI Conference on Artificial Intelligence

      Volume: 39 Pages: 19749~19757

    • DOI

      10.1609/aaai.v39i18.34175

    • Peer Reviewed / Open Access
  • [Journal Article] A Logical Analysis of Hanabi2025

    • Author(s)
      Perrotin Elise
    • Journal Title

      Proceedings of the AAAI Conference on Artificial Intelligence

      Volume: 39 Pages: 15118~15125

    • DOI

      10.1609/aaai.v39i14.33658

    • Peer Reviewed / Open Access
  • [Journal Article] A Computationally Grounded Framework for Cognitive Attitudes2025

    • Author(s)
      De Lima Tiago, Lorini Emiliano, Perrotin Elise, Schwarzentruber Francois
    • Journal Title

      Proceedings of the AAAI Conference on Artificial Intelligence

      Volume: 39 Pages: 14858~14866

    • DOI

      10.1609/aaai.v39i14.33629

    • Peer Reviewed / Open Access / Int'l Joint Research
  • [Journal Article] Neuro-Symbolic Contrastive Learning for Cross-domain Inference2025

    • Author(s)
      Mingyue Liu, Ryo Ueda, Zhen Wan, Katsumi Inoue, Chris G. Willcocks
    • Journal Title

      Proceedings of the 40th International Conference on Logic Programming (ICLP 2024), Electronic Proceedings in Theoretical Computer Science

      Volume: 416 Pages: 78~94

    • DOI

      10.4204/EPTCS.416.6

    • Peer Reviewed / Open Access / Int'l Joint Research
  • [Journal Article] On Automatic Generation of?Escher-Like Metamorphosis2025

    • Author(s)
      Nakamatsu Shunsuke, Sakama Chiaki
    • Journal Title

      Multi-disciplinary Trends in Artificial Intelligence - 17th International Conference, MIWAI 2024, Pattaya, Thailand, November 11-15, 2024, Proceedings, Part I, LNAI

      Volume: 15431 Pages: 419~431

    • DOI

      10.1007/978-981-96-0692-4_34

    • Peer Reviewed
  • [Journal Article] Generating Global and Local Explanations for Tree-Ensemble Learning Methods by Answer Set Programming2024

    • Author(s)
      Takemura Akihiro, Inoue Katsumi
    • Journal Title

      Theory and Practice of Logic Programming

      Volume: 24 Pages: 973~1010

    • DOI

      10.1017/S1471068424000401

    • Peer Reviewed / Open Access
  • [Journal Article] A differentiable first-order rule learner for inductive logic programming2024

    • Author(s)
      Gao Kun, Inoue Katsumi, Cao Yongzhi, Wang Hanpin
    • Journal Title

      Artificial Intelligence

      Volume: 331 Pages: 104108~104108

    • DOI

      10.1016/j.artint.2024.104108

    • Peer Reviewed / Int'l Joint Research
  • [Journal Article] Future-proofing class-incremental learning2024

    • Author(s)
      Jodelet Quentin, Liu Xin, Phua Yin Jun, Murata Tsuyoshi
    • Journal Title

      Machine Vision and Applications

      Volume: 36 Pages: -

    • DOI

      10.1007/s00138-024-01635-y

    • Peer Reviewed / Open Access
  • [Journal Article] Floor-Field-Guided Neural Model for Crowd Counting2024

    • Author(s)
      Habara Takehiro, Kojima Ryosuke
    • Journal Title

      IEEE Access

      Volume: 12 Pages: 154888~154900

    • DOI

      10.1109/ACCESS.2024.3483252

    • Peer Reviewed / Open Access
  • [Journal Article] Large Neighborhood Prioritized Search for Combinatorial Optimization with Answer Set Programming2024

    • Author(s)
      Sugimori Irumi、Inoue Katsumi、Nabeshima Hidetomo、Schaub Torsten、Soh Takehide、Tamura Naoyuki、Banbara Mutsunori
    • Journal Title

      Proceedings of the 21st International Conference on Knowledge Representation and Reasoning (KR-2024; Hanoi, Vietnam, November 2-8, 2024)

      Volume: - Pages: 765~776

    • DOI

      10.24963/kr.2024/72

    • Peer Reviewed / Open Access / Int'l Joint Research
  • [Journal Article] Abductive Reasoning in a Paraconsistent Framework2024

    • Author(s)
      Bienvenu Meghyn, Inoue Katsumi, Kozhemiachenko Daniil
    • Journal Title

      Proceedings of the 21st International Conference on Knowledge Representation and Reasoning (KR-2024; Hanoi, Vietnam, November 2-8, 2024)

      Volume: - Pages: 134~144

    • DOI

      10.24963/kr.2024/13

    • Peer Reviewed / Open Access / Int'l Joint Research
  • [Journal Article] Linear Algebraic Partial Evaluation of Logic Programs2024

    • Author(s)
      Nguyen Tuan, Inoue Katsumi, Sakama Chiaki
    • Journal Title

      Proceedings of the 36th IEEE International Conference on Tools with Artificial Intelligence (ICTAI 2024; Herndon, VA, USA, October 30 - November 1, 2024)

      Volume: - Pages: 355~362

    • DOI

      10.1109/ICTAI62512.2024.00059

    • Peer Reviewed
  • [Journal Article] Differentiable Logic Programming for Distant Supervision2024

    • Author(s)
      Takemura Akihiro, Inoue Katsumi
    • Journal Title

      Proceedings of the 27th European Conference on Artificial Intelligence (ECAI-2024; Santiago de Compostela, Spain, October 19-24, 2024), Frontiers in Artificial Intelligence and Applications

      Volume: 392 Pages: 1301-1308

    • DOI

      10.3233/faia240628

    • Peer Reviewed / Open Access
  • [Journal Article] Variable Assignment Invariant Neural Networks for Learning Logic Programs2024

    • Author(s)
      Phua Yin Jun, Inoue Katsumi
    • Journal Title

      Neural-Symbolic Learning and Reasoning (NeSy 2024), LNAI

      Volume: 14979 Pages: 47~61

    • DOI

      10.1007/978-3-031-71167-1_3

    • Peer Reviewed
  • [Journal Article] Linearly Structured World Representations in Maze-Solving Transformers2024

    • Author(s)
      Michael Ivanitskiy, Alexander Spies, Tilman Rauker, Guillaume Corlouer, Christopher Mathwin, Lucia Quirke, Can Rager, Rusheb Shah, Dan Valentine, Cecilia Diniz Behn, Katsumi Inoue, Samy Wu Fung
    • Journal Title

      Proceedings of UniReps: the First Workshop on Unifying Representations in Neural Models). Proceedings of Machine Learning Research

      Volume: 243 Pages: 133-143

    • Peer Reviewed / Open Access / Int'l Joint Research
  • [Journal Article] Argument and Belief2024

    • Author(s)
      Sakama Chiaki
    • Journal Title

      The 10th International Conference on Computational Models of Argument (COMMA 2024), Frontiers in Artificial Intelligence and Applications

      Volume: 388 Pages: 241-252

    • DOI

      10.3233/FAIA240325

    • Peer Reviewed / Open Access
  • [Journal Article] Relative Change-Reluctance in Iterated Belief Revision2024

    • Author(s)
      Perrotin Elise, Schwind Nicolas
    • Journal Title

      The 21st Pacific Rim International Conference on Artificial Intelligence (PRICAI 2024), LNAI

      Volume: 15285 Pages: 264~276

    • DOI

      10.1007/978-981-96-0128-8_23

    • Peer Reviewed
  • [Journal Article] Reconsidering Stochastic Policy Gradient Methods for?Traffic Signal Control2024

    • Author(s)
      Kato Masahiro, Kojima Ryosuke
    • Journal Title

      Advances and Trends in Artificial Intelligence. Theory and Applications (IEA/AIE 2024), LNAI

      Volume: 14748 Pages: 442~453

    • DOI

      10.1007/978-981-97-4677-4_36

    • Peer Reviewed
  • [Presentation] Differentiable Rule Induction from Raw Sequence Inputs2025

    • Author(s)
      Kun Gao, Katsumi Inoue, Yongzhi Cao, Hanpin Wang, Yang Feng
    • Organizer
      The 13th International Conference on Learning Representations (ICLR 2025; Singapore, April 24-28, 2025)
    • Int'l Joint Research
  • [Presentation] Transformers Use Causal World Models in Maze-Solving Tasks2025

    • Author(s)
      Alexander F. Spies, William Edwards, Michael Ivanitskiy, Adrians Skapars, Tilman Rauker, Katsumi Inoue, Alessandra Russo, Murray Shanahan
    • Organizer
      ICLR 2025 Workshop on World Models: Understanding, Modelling and Scaling
    • Int'l Joint Research
  • [Presentation] Backtracking Enabled Transformers2025

    • Author(s)
      Yin Jun Phua, Tony Ribeiro, Katsumi Inoue
    • Organizer
      The 9th International Conference on Machine Learning and Soft Computing (ICMLSC 2025)
    • Int'l Joint Research
  • [Presentation] Linear Algebraic Partial Evaluation of Logic Programs2024

    • Author(s)
      Tuan Nguyen Quoc, Katsumi Inoue, Chiaki Sakama
    • Organizer
      The 22nd International Workshop on Nonmonotonic Reasoning (NMR 2024)
    • Int'l Joint Research
  • [Presentation] A Constraint-Based Visual Dataset Generator2024

    • Author(s)
      Thomas Eiter, Nelson Higuera, Katsumi Inoue, Sota Moriyama
    • Organizer
      The 7th Workshop on Trends and Applications of Answer Set Programming (TAASP 2024)
    • Int'l Joint Research
  • [Presentation] Linear Algebraic Approaches to Logic Programming2024

    • Author(s)
      Katsumi Inoue
    • Organizer
      The 40th International Conference on Logic Programming (ICLP 2024)
    • Int'l Joint Research / Invited
  • [Presentation] Algebraic Logic Programming and Learning2024

    • Author(s)
      Katsumi Inoue
    • Organizer
      The 4th International Joint Conference on Learning & Reasoning (IJCLR 2024)
    • Int'l Joint Research / Invited
  • [Presentation] Learning From Interpretation Transitions with Unknowns2024

    • Author(s)
      Tony Ribeiro, Maxime Folschette, Morgan Magnin, Kotaro Okazaki, Lo Kuo-Yen, Antoine Roquilly, Jeremie Poschmann, Katsumi Inoue
    • Organizer
      The 4th International Joint Conference on Learning & Reasoning (IJCLR 2024)
    • Int'l Joint Research
  • [Presentation] A differentiable first-order rule learner for inductive logic programming2024

    • Author(s)
      Kun Gao, Katsumi Inoue, Yongzhi Cao, Hanpin Wang
    • Organizer
      The 33rd International Joint Conference on Artificial Intelligence (IJCAI 2024)
    • Int'l Joint Research / Invited
  • [Presentation] Variable Assignment Invariant Neural Network for Learning Logic Program2024

    • Author(s)
      Yin Jun Phua, Katsumi Inoue
    • Organizer
      The 8th International Workshop on Symbolic-Neural Learning (SNL 2024)
    • Int'l Joint Research
  • [Presentation] Data-driven Analysis of Domain Specificity for Explainable Session-based Recommendation System2024

    • Author(s)
      Kotaro Okazaki, Tony Ribeiro, Kuo-Yen Lo, Junichi Sakuma, Katsumi Inoue
    • Organizer
      2024年度人工知能学会全国大会(第38回)国際セッション
    • Int'l Joint Research
  • [Presentation] マルチラベル物体認識への制約知識の導入とROAD-Rへの適用2024

    • Author(s)
      森山 総太, 渡邉 晃司, 井上 克巳, 竹村 彰浩
    • Organizer
      2024年度人工知能学会全国大会(第38回)
  • [Presentation] 微分可能な論理プログラムによる Distant Supervision2024

    • Author(s)
      竹村 彰浩, 井上 克巳
    • Organizer
      2024年度人工知能学会全国大会(第38回)
  • [Presentation] 論理制約を考慮したテーブルデータを対象とした予測モデル構築フレームワーク2024

    • Author(s)
      尾上圭介, 小島諒介
    • Organizer
      2024年度人工知能学会全国大会(第38回)
  • [Presentation] Algebraic Connection between Logic Programming and Machine Learning2024

    • Author(s)
      Katsumi Inoue
    • Organizer
      The 17th International Symposium on Functional and Logic Programming (FLOPS 2024)
    • Int'l Joint Research / Invited
  • [Remarks] 井上研究室ホームページ

    • URL

      http://research.nii.ac.jp/il/

  • [Remarks] 坂間千秋

    • URL

      http://web.wakayama-u.ac.jp/~sakama

URL: 

Published: 2025-12-26  

Information User Guide FAQ News Terms of Use Attribution of KAKENHI

Powered by NII kakenhi