Machine learning assisted interatomic and electronic structure models for molecular simulation

We introduce a machine learning (ML)-based framework for building interatomic and electronic structure models following two general principles: 1) ML-based models should respect important physical constraints in a faithful and adaptive way; 2) to build truly reliable models, efficient algorithms are needed to explore relevant physical space and construct optimal training data sets. Two examples are given: 1) DeePMD, an end-to-end symmetry-preserving model for efficient molecular dynamics with ab initio accuracy; 2) DeePKS, a chemically accurate and widely-applicable electronic structure model within the framework of generalized Kohn-Sham density functional theory. If time permits, we will also present our efforts on developing related software packages and high-performance computing schemes, which have now been widely used worldwide by experts and practitioners in the molecular and materials simulation community.