Machine learning for decision making based on complex structured data

2023 Fiscal Year Final Research Report

• Project/Area Number: 20H04244
• Research Category: Grant-in-Aid for Scientific Research (B)
• Allocation Type: Single-year Grants
• Section: 一般
• Review Section: Basic Section 61030:Intelligent informatics-related
• Research Institution: Kyoto University
• Principal Investigator: Kashima Hisashi 京都大学, 情報学研究科, 教授 (80545583)
• Project Period (FY): 2020-04-01 – 2024-03-31
• Keywords: 機械学習 / 人工知能 / 因果推論

Outline of Final Research Achievements

First, with the aim of expanding the applicability of machine learning, we improved the performance of deep learning methods for graph-structured data, and developed models that are more expressive than conventional models and effective learning methods for them. In addition, with the aim of expanding the applicability of data-driven decision making, we developed causal effect estimation methods in situations where confounding variables are unknown, applied causal effect estimation to the field of chemistry, and developed predictive modeling methods for small data. Furthermore, we combined graph deep learning and causal inference to develop causal effect estimation for interventions with graph structure and causal effect estimation methods on graphs.

Free Research Field

機械学習

Academic Significance and Societal Importance of the Research Achievements

グラフ構造データは、ソーシャルネットワーク、分子構造、交通網など多様な分野で見られる。高い性能をもつグラフ深層学習モデルの開発、さらには深層因果推論手法との融合によって、これらの分野におけるより高度な意思決定を可能とし、新薬の発見、交通最適化、社会的ダイナミクスの理解など様々な実世界応用の可能性をもつ。