辦公平臺 | 所內郵箱 | ARP | English
當前位置: 首頁 > 科研成果 > 學術活動 > 學術報告
[理論室學術報告] Enhanced sampling and free energy calculation in high-dimensional coarse-grained space
時間: 2019年07月04日 09:30
地點: M830
報告人: Linfeng Zhang

PACM, Princeton University

Abstract: We consider a learning problem in statistical physics: given an ensemble \rho(x)=e^{-\beta H(x)}/Z and a coarse-graining procedure y=y(x) that maps x to a reduced set of variables y, we need to model the free energy surface F(y) = - (1/\beta) \log \int dx\rho(x)\delta(y-y(x)). This problem is relevant to situations in several different fields, and we will discuss examples from statistical lattice models [1] and molecular dynamics [2,3]. A common challenge to these examples is the so-called curse of dimensionality, i.e., when y is in a high-dimensional space, three non-trivial issues will be intertwined with each other: the representation of the free energy surface, the optimization of the parameters in the representation, and the exploration of relevant phase space points. We will see how methods in Refs. [1-3] address these issues.

[1] Yantao Wu and Roberto Car. "Variational approach to monte carlo renormalization group." Physical review letters 119.22 (2017): 220602.

[2] Linfeng Zhang, Han Wang, and Weinan E. "Reinforced dynamics for enhanced sampling in large atomic and molecular systems." The Journal of chemical physics 148.12 (2018): 124113.

[3] Linfeng Zhang, De-Ye Lin, Han Wang, Roberto Car, and Weinan E. "Active learning of uniformly accurate interatomic potentials for materials simulation." Physical Review Materials. 2019 Feb 25;3(2):023804.

Contact: Lei Wang, 9853


中科院logo
海码论坛香港赛马会