报告题目：Deep learning for multi-scale molecular modeling
We introduce a series of deep learning based methods and applications in the context of multi-scale molecular modeling [1-6]. Examples include 1) potential energy surface and molecular dynamics [1-3]; 2) free energy surface and enhanced sampling [4-5]; and - if time permits - 3) many-electron wavefunction and variational Monte Carlo . From these examples, we summarize two general principles: 1) model construction and data exploration interplay in an iterative way; 2) the key component at different scales is how deep learning based models respect physical constraints.
 Linfeng Zhang, Jiequn Han, Han Wang, Roberto Car, and Weinan E, Physical Review Letters 120 (2018): 143001.
 Linfeng Zhang, Jiequn Han, Han Wang, Wissam Saidi, Roberto Car, and Weinan E, Advances in Neural Information Processing Systems (2018): 4441-4451.
 Linfeng Zhang, De-Ye Lin, Han Wang, Roberto Car, and Weinan E, Phys. Rev. Materials 3 (2019): 023804.
 Linfeng Zhang, Han Wang, and Weinan E, J. Chem. Phys.148 (2018):124113.
 Linfeng Zhang, Jiequn Han, Han Wang, Roberto Car, and Weinan E, J. Chem. Phys.149 (2018):034101.
 Jiequn Han, Linfeng Zhang, Weinan E, arxiv: 1807.07014.
Linfeng Zhang graduated from Peking University in 2016. He is now a graduate student in the Program in Applied and Computational Mathematics (PACM), Princeton University, working with Profs. Roberto Car and Weinan E. Linfeng is interested in various mathematical and physical problems originated from different disciplines of sciences. Most recently he has been focusing on developing deep learning based models for quantum chemistry, molecular dynamics, as well as enhanced sampling. Linfeng is one of the main developers of DeePMD-kit, a very popular deep learning based open-source software for molecular dynamics.