人工智能培训

搜索

人工智能论文:分子动力学的变分粗粒度(Variational Coarse-Graining for Molecular Dynamics)

[复制链接]
comceo 发表于 2018-12-7 11:35:36 | 显示全部楼层 |阅读模式
comceo 2018-12-7 11:35:36 230 0 显示全部楼层
人工智能论文:分子动力学的变分粗粒度(Variational Coarse-Graining for Molecular Dynamics)分子动力学模拟提供了对凝聚相中材料的微观行为的理论洞察,并且作为预测工具,实现了新化合物的计算设计。然而,由于材料中热力学和动力学现象所涉及的大时空尺度,原子模拟通常在计算上是不可行的。粗粒化方法允许通过减少模拟的维度来模拟更大的系统,并通过减少快速运动来传播更长的时间步长。粗粒度涉及两个耦合学习问题;定义从全原子到简化表示的映射,并在粗粒度坐标上参数化哈密顿量。多元统计力学方法已经解决了后者,但前者通常是基于化学直觉的手动调整过程。在这里,我们提出了一个基于自动编码器的优化框架,同时学习这两个任务。 Autograin自动学习全原子和简化表示之间的最佳映射,使用重建损失促进粗粒度变量的学习。此外,应用力匹配方法来变分地确定粗粒度势能函数。该程序在许多模型系统上进行测试,包括单分子和体相周期模拟。
Molecular dynamics simulations provide theoretical insight into themicroscopic behavior of materials in condensed phase and, as a predictive tool,enable computational design of new compounds.However, because of the largetemporal and spatial scales involved in thermodynamic and kinetic phenomena inmaterials, atomistic simulations are often computationally unfeasible.Coarse-graining methods allow simulating larger systems, by reducing thedimensionality of the simulation, and propagating longer timesteps, byaveraging out fast motions.Coarse-graining involves two coupled learningproblems;defining the mapping from an all-atom to a reduced representation,and parametrizing a Hamiltonian over coarse-grained coordinates.Multiplestatistical mechanics approaches have addressed the latter, but the former isgenerally a hand-tuned process based on chemical intuition.Here we presentAutograin, an optimization framework based on auto-encoders to learn both taskssimultaneously.Autograin automatically learns the optimal mapping betweenall-atom and reduced representation, using the reconstruction loss tofacilitate the learning of coarse-grained variables.In addition, aforce-matching method is applied to variationally determine the coarse-grainedpotential energy function.This procedure is tested on a number of modelsystems including single-molecule and bulk-phase periodic simulations.人工智能论文:分子动力学的变分粗粒度(Variational Coarse-Graining for Molecular Dynamics) CCjjr1i6dVj1ezcc.jpg
URL地址:https://arxiv.org/abs/1812.02706     ----pdf下载地址:https://arxiv.org/pdf/1812.02706    ----人工智能论文:分子动力学的变分粗粒度(Variational Coarse-Graining for Molecular Dynamics)
回复

使用道具 举报

您需要登录后才可以回帖 登录 | 立即注册

本版积分规则 返回列表 发新帖

comceo当前离线
新手上路

查看:230 | 回复:0

快速回复 返回顶部 返回列表