[18a-B01-1~4] New material science and its practical application created by informatics

[18a-B01-2] Materials Informatics: Opportunities and Challenges for Inorganic Materials

〇Yuzuru Tanaka^1,2,3, Jun Fujima¹, Keisuke Takahashi^4,1 (1.NIMS MaDIS, 2.CROSS, 3.Faculty of Science, University Alberta, 4.Faculty of Science, Hokkaidou University)

Keywords:materarials informatics, functional inorganic crystal materials, heterogeneity of data

Materials informatics is now moving from the age of success with simple application of machine learning to the age of requiring new mathematical modeling of materials suitable for machine learning. In search for functional inorganic materials, the target data set is quite heterogeneous and requires appropriate data segmentation before regression analysis. Each homogeneous data set after the segmentation becomes very small, and it is extremely difficult to define a small number of appropriate descriptors to avoid overfitting. We will discuss how to cope with these issues.