[14p-A205-1~11] 23.1 Joint Session N "Informatics"

[14p-A205-5] XRD generation for unknown samples in Auto-Encoder

〇Keishu Uchimura¹, Masao Yano², Hiroyuki Kimoto², Kenta Hongo^1,3,4, Ryo Maezono¹ (1.JAIST, 2.TOYOTA, 3.NIMS, 4.PRESTO, JST)

Keywords:Machine learning, XRD, Ab initio calculation

In this study, we generate XRD from given composition by learning the relationship between composition and XRD using a neural network. To prepare the XRD data for learning, we performed ab initio calculation. We constructed the auto-encoder learns to export the XRD from the XRD. It was found that the feature space corresponds to the composition. Since each point on the feature space has a meaning of the composition, the decoder part is capable of generating the XRD from the information of the composition.