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

Explanation-guided Machine Learning Model Development

Research Project

Project/Area Number 20K19860
Research Category

Grant-in-Aid for Early-Career Scientists

Allocation TypeMulti-year Fund
Review Section Basic Section 61030:Intelligent informatics-related
Research InstitutionOsaka University

Principal Investigator

Hara Satoshi  大阪大学, 産業科学研究所, 准教授 (40780721)

Project Period (FY) 2020-04-01 – 2024-03-31
Project Status Completed (Fiscal Year 2023)
Budget Amount *help
¥4,160,000 (Direct Cost: ¥3,200,000、Indirect Cost: ¥960,000)
Fiscal Year 2022: ¥1,170,000 (Direct Cost: ¥900,000、Indirect Cost: ¥270,000)
Fiscal Year 2021: ¥1,300,000 (Direct Cost: ¥1,000,000、Indirect Cost: ¥300,000)
Fiscal Year 2020: ¥1,690,000 (Direct Cost: ¥1,300,000、Indirect Cost: ¥390,000)
Keywords機械学習 / 深層学習 / 説明可能AI / 人工知能
Outline of Research at the Start

実用に供するレベルの精度の高い機械学習モデルを作ることは必ずしも容易ではなく、開発者の熟練度によって出来上がるモデルの精度には大きな開きがある。本研究では説明可能AI(Explainable AI; XAI)の技術を基盤に、開発者により良いモデルの構築方法について適切にアドバイスする仕組み「説明駆動モデル開発」の研究に取り組む。本研究では特に「説明駆動モデル開発」の基盤技術として、XAIの技術を機械学習モデルの開発へと拡張することで、(i)良い学習方法、および(ii)悪いモデルの修理方法、を推定して開発者へとフィードバックする仕組みの確立を目指す。

Outline of Final Research Achievements

In this reserach project, we focused on the development of "Explanation-guided model development," which provides appropriate advice to developers on how to construct better machine learning models, leveraging Explainable AI technology as its foundation. Creating highly accurate machine learning models at a level suitable for practical use is not always straightforward, and there can be significant variations in the accuracy of models produced depending on the skill level of the developers.
In this project, we developed methods for data cleansing to improve model performance, its extesion for similarity-based explanation, as well as model correction techniques based on partial model explanation.

Academic Significance and Societal Importance of the Research Achievements

研究成果の学術的意義としては「説明駆動モデル開発という新たな機械学習モデルの開発の仕組みの提案」、そして「XAI技術のさらなる発展による機械学習モデルの解釈性の向上」があげられる。これらにより、モデルを効率的に改善する方法や、モデルがなぜ特定の予測や判断を下したのかを理解することが容易になり、モデルへの信頼性向上が期待される。
研究成果の社会的意義としては「高性能なモデルが効率的に開発可能になることで、機械学習モデルの社会的な活用がより一層進む」ことがあげられる。

Report

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

    (9 results)

All 2023 2022 2021 2020 Other

All Int'l Joint Research (3 results) Presentation (6 results) (of which Int'l Joint Research: 4 results)

  • [Int'l Joint Research] Universite du Quebec a Montreal(カナダ)

    • Related Report
      2022 Research-status Report
  • [Int'l Joint Research] Universite du Quebec a Montreal(カナダ)

    • Related Report
      2021 Research-status Report
  • [Int'l Joint Research] Universite du Quebec a Montreal(カナダ)

    • Related Report
      2020 Research-status Report
  • [Presentation] Rule Mining for Correcting Classification Models2023

    • Author(s)
      Hirofumi Suzuki, Hiroaki Iwashita, Takuya Takagi, Yuta Fujishige, Satoshi Hara
    • Organizer
      23rd IEEE International Conference on Data Mining
    • Related Report
      2023 Annual Research Report
    • Int'l Joint Research
  • [Presentation] 決定損失の期待値と分散を用いた分類モデルの較正2023

    • Author(s)
      宗近康平, 原聡
    • Organizer
      2023年度人工知能学会全国大会
    • Related Report
      2023 Annual Research Report
  • [Presentation] Explainable and Local Correction of Classification Models Using Decision Trees2022

    • Author(s)
      Hirofumi Suzuki, Hiroaki Iwashita, Takuya Takagi, Keisuke Goto, Yuta Fujishige, Satoshi Hara
    • Organizer
      The 36th AAAI Conference on Artificial Intelligence
    • Related Report
      2022 Research-status Report
    • Int'l Joint Research
  • [Presentation] Evaluation of Similarity-based Explanations2021

    • Author(s)
      Kazuaki Hanawa, Sho Yokoi, Satoshi Hara, Kentaro Inui
    • Organizer
      The 9th International Conference on Learning Representations (ICLR'21)
    • Related Report
      2021 Research-status Report
    • Int'l Joint Research
  • [Presentation] Characterizing the risk of fairwashing2021

    • Author(s)
      Ulrich Aivodji, Hiromi Arai, Sebastien Gambs, Satoshi Hara
    • Organizer
      Neural Information Processing Systems 34 (NeurIPS'21)
    • Related Report
      2021 Research-status Report
    • Int'l Joint Research
  • [Presentation] Data Cleansing for Reinforcement Learning with Least Squares Temporal Difference2020

    • Author(s)
      藩丹青
    • Organizer
      第23回情報論的学習理論ワークショップ(IBIS2020)
    • Related Report
      2020 Research-status Report

URL: 

Published: 2020-04-28   Modified: 2025-01-30  

Information User Guide FAQ News Terms of Use Attribution of KAKENHI

Powered by NII kakenhi