[8a-N302-1~11] 18.1 Materials Informatics

10:00 AM - 10:15 AM JST(1:00 AM - 1:15 AM UTC)

[8a-N302-5] Interpreting the Decision-Making Rationale for Autonomous Material Discovery Using

〇KOHEI KATSUDA¹, Yoshida Naoki¹, Iwabuchi Yutaro², Iwasaki Yuma², Igarashi Yasuhiko^1,2 (1.Tsukuba Univ., 2.NIMS)

Keywords:

bayesian optimization

In the field of materials discovery, there is growing interest in autonomous materials discovery systems that utilize Bayesian optimization. However, the rationale behind the proposed experimental conditions tends to become a “black box,” posing challenges when interpreting the discovery results as insights for materials design. In this study, we apply SHAP analysis to the discovery model and quantitatively evaluate the material features that contributed to candidate selection, thereby visualizing the decision-making process and verifying its interpretability.