← 研踪 PaperTrailarXiv:2606.04100 · v2
Stein Kernelized Molecular Dynamics for Active Learning of Interatomic Potentials
首次提交 2026-06-03 02:07
版本更新 2026-11-17 16:56
雷达观测 2026-11-18 19:42
摘要
Machine learning interatomic potentials (MLIPs) enable efficient and accurate atomistic simulations but depend critically on the quality and diversity of the training data. We introduce Stein kernelized molecular dynamics (SKMD), an enhanced sampling method that uses interacting particle dynamics to acquire informative training configurations for the fine-tuning of MLIPs. SKMD corresponds to a stochastic variant of Stein variational gradient descent that is adapted for molecular dynamics by incorporating asynchronous particle updates and a kernel of global atomic descriptors, which provides a symmetry-aware measure of configurational similarity. Unlike other enhanced samplers used in molecular dynamics, SKMD maintains the Boltzmann distribution as a stationary distribution of its dynamics in the appropriate asymptotic limit. This property enforces a balance between the exploration of diverse configurations and attraction toward high-probability regions of the energy landscape. We further propose an approach to efficient online data acquisition using an adaptive stopping criterion that selects non-redundant training data over the course of simulation. We demonstrate SKMD for the active learning of a neural network model of the Müller--Brown potential, a MACE interatomic potential for alanine dipeptide, and a MACE foundation model with a DFT oracle for predicting the Pandey reconstruction of the Si(111) surface. Compared to active learning baselines, our method achieves higher model accuracy in fewer training iterations.
观测到的版本
相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。
v2 · 2026-11-17 16:56 · 最新观测版本
Stein Kernelized Molecular Dynamics for Active Learning of Interatomic Potentials
Machine learning interatomic potentials (MLIPs) enable efficient and accurate atomistic simulations but depend critically on the quality and diversity of the training data. We introduce Stein kernelized molecular dynamics (SKMD), an enhanced sampling method that uses interacting particle dynamics to acquire informative training configurations for the fine-tuning of MLIPs. SKMD corresponds to a stochastic variant of Stein variational gradient descent that is adapted for molecular dynamics by incorporating asynchronous particle updates and a kernel of global atomic descriptors, which provides a symmetry-aware measure of configurational similarity. Unlike other enhanced samplers used in molecular dynamics, SKMD maintains the Boltzmann distribution as a stationary distribution of its dynamics in the appropriate asymptotic limit. This property enforces a balance between the exploration of diverse configurations and attraction toward high-probability regions of the energy landscape. We further propose an approach to efficient online data acquisition using an adaptive stopping criterion that selects non-redundant training data over the course of simulation. We demonstrate SKMD for the active learning of a neural network model of the Müller--Brown potential, a MACE interatomic potential for alanine dipeptide, and a MACE foundation model with a DFT oracle for predicting the Pandey reconstruction of the Si(111) surface. Compared to active learning baselines, our method achieves higher model accuracy in fewer training iterations.
查看此版本解读 →查看此版本原文 ↗