[18a-A401-1~9] 23.1 Joint Session N "Informatics"

[18a-A401-5] Peak relevancy in XRD identified by auto-encoder technique

〇Ryo Maezono¹, Keishu Utimula², Kenta Hongo³ (1.JAIST, Info. Sci., 2.JAIST, Mater. Sci., 3.JAIST, RCACI)

Keywords:Auto-encoder, XRD, Machine-learning

We developed a scheme to identify which XRD peaks are relevant to characterize a systematically varying quantity in a series of samples by using an auto-encoder technique [1,2]. Individual XRD patterns are projected onto a single point in a two-dimensional feature space constructed using the encoder. If a point is significantly shifted when a peak of interest is masked, then the peak is relevant for characterizing the pattern represented by the point on the space. In this manner, we can quantitatively formulate the relevance of a peak.