[18a-A21-1~11] 23.1 Joint Session N "Informatics"

9:30 AM - 9:45 AM JST(12:30 AM - 12:45 AM UTC)

[18a-A21-3] Geometric concept learning of crystal structures

〇Keisuke Ozawa¹, Teppei Suzuki¹, Shunsuke Tonogai², Tomoya Itakura² (1.DENSO IT Lab., 2.DENSO CORP.)

Keywords:

Inorganic materials,Crystal structure,Contrastive learning

We present a machine learning method for embedding geometric concepts of crystal structures. Using crystal graphs and texts describing geometric concepts, we conducted a contrastive learning between the graphs and texts. We discuss the results of similar crystal structure search and how it improves the classification accuracy of crystal structures using few training data.