Limits of machine learning models for atomic simulations in high dimensions
On the Theoretical Limits of Learning Unconstrained Interatomic Potentials
Machine Learning
Summary
Machine-learned models are used to speed up simulations of how atoms move by predicting their energies and forces. This paper shows that when these models have no specific physical constraints, their predictions may look accurate but can still lead to very wrong behaviors in actual simulations, especially as the number of atoms increases. Two key problems arise: simulations get stuck for a very long time, or they settle into incorrect stable states. The authors also provide conditions that, if met, guarantee accurate and fast simulations.
What this means in practice
- •For computational chemists: Assess risks of inaccurate sampling when using unconstrained ML models for simulating atomic dynamics in complex molecules.
- •For materials simulation engineers: Design molecular dynamics simulations that include energy constraints to ensure accurate and efficient sampling of atomic configurations.
A theory result. No direct application yet.
Authors
Prajwal Ananth, Shuwen Yue
Abstract
Machine-learned interatomic potentials (MLIPs) accelerate molecular dynamics (MD) simulations by learning the energies and forces that govern atomic motion. Recent work has shown that relaxing physical inductive biases can improve computational efficiency and predictive accuracy, raising the question of whether energy-and-force matching alone is sufficient to ensure accurate sampling. We show that, for unconstrained hypothesis classes, the relationship between prediction error and sampling accuracy is exponentially ill-conditioned in dimension. In particular, prediction errors that decay exponentially with dimension can coexist with nearly maximal total variation distance between the learned and target distributions. We identify two failure mechanisms. The first is kinetic trapping, where relaxation to an accurate learned equilibrium requires exponentially long times, even when simulations are initialized from training configurations at which energies and forces are matched exactly. The second is equilibrium mismatch, where the learned stationary distribution itself remains far from the target. Under explicit assumptions, we show that equilibrium mismatch can persist even in local, symmetry-preserving models, where small per-atom errors coexist with a distributional discrepancy that approaches its maximum as system size increases. Finally, we establish global energy-control conditions that guarantee equilibrium accuracy and identify additional conditions required for rapid relaxation.