Neural networks-based variationally enhanced sampling
Neural networks-based variationally enhanced sampling
复制标题
DOI：
10.1073/pnas.1907975116
复制
发表时间：
2019-09-03
影响因子：
11.1
通讯作者：
Parrinello, Michele
中科院分区：
文献类型：
--
作者：
Bonati, Luigi;Zhang, Yue-Yu;Parrinello, Michele
复杂自由能表面的采样是现代原子模拟方法的主要挑战之一。在这样的表面中存在动力学瓶颈，常常使得直接方法无用。一种流行的策略是识别少量关键的集体变量，并引入能够支持其波动的偏差势，以加速采样。在这里，我们建议将机器学习技术与最近的变分增强采样方法结合使用[O。Valsson，M. Parrinello，Phys. Rev. Lett. 113，090601（2014）]，以确定这种潜力。这是通过将偏差表示为神经网络来实现的。在变分学习计划，旨在最大限度地减少适当的功能的参数。这就需要开发一种更有效的最小化技术。神经网络的表现力允许表示快速变化的自由能表面，消除边界效应伪影，并允许处理几个集体变量。
Sampling complex free-energy surfaces is one of the main challenges of modern atomistic simulation methods. The presence of kinetic bottlenecks in such surfaces often renders a direct approach useless. A popular strategy is to identify a small number of key collective variables and to introduce a bias potential that is able to favor their fluctuations in order to accelerate sampling. Here, we propose to use machine-learning techniques in conjunction with the recent variationally enhanced sampling method [O. Valsson, M. Parrinello, Phys. Rev. Lett. 113, 090601 (2014)] in order to determine such potential. This is achieved by expressing the bias as a neural network. The parameters are determined in a variational learning scheme aimed at minimizing an appropriate functional. This required the development of a more efficient minimization technique. The expressivity of neural networks allows representing rapidly varying free-energy surfaces, removes boundary effects artifacts, and allows several collective variables to be handled.