[19p-C601-1~15] 23.1 Joint Session N "Informatics"

4:45 PM - 5:00 PM JST(7:45 AM - 8:00 AM UTC)

[19p-C601-12] Bayesian optimization framework for high ionic conductivity material exploration

〇(DC)Yuki Sakishita¹, Yibin Xu², Koji Hukushima^1,3 (1.Basic Sci., Univ. of Tokyo, 2.NIMS, 3.Komaba Inst. for Sci., Univ. of Tokyo)

Keywords:

Bayesian optimization,materials informatics,ion conductor

We propose a Bayesian optimization framework using machine learning models for the exploration of highly ionic conductive materials. A Bayesian prediction model with low-dimensional descriptors reflecting the periodic table is designed and a Bayesian optimization method using the evaluation of the prediction uncertainty of this model enables efficient material exploration. Based on these, a stochastic search that maximize the acquisition function in a multidimensional parameter space, and a framework that facilitates parallel experiments by simultaneously proposing multiple candidate materials are presented.