Symbolic Regression in Materials Informatics: Applications and Challenges

Back

Symbolic Regression in Materials Informatics: Applications and Challenges 

Runhai Ouyang

Principal Investigator at the Materials Genome Institute, Shanghai University

Symbolic regression (SR) is a key artificial intelligence method for generating interpretable descriptors in materials and chemistry informatics. Many SR methods have been developed in recent years, including the sure independence screening and sparsifying operator (SISSO) [1]. While SR has demonstrated great success in accelerated materials discovery, major challenges remain in representing complex atomistic structure for compact expressions. In this talk, I will review the current strategies to bypass this difficulty for SR application in materials and chemistry informatics based on the method SISSO, and present my perspective on future development in algorithm. 

[1] R. Ouyang, S. Curtarolo, E. Ahmetcik, M. Scheffler, L. M. Ghiringhelli, Phys. Rev. Mater. 2, 083802 (2018).