Understanding and Mitigating Distribution Shifts For Machine Learning Force Fields

TL;DR

This paper investigates how machine learning force fields (MLFFs) generalize across diverse chemical spaces, identifies challenges due to distribution shifts, and proposes test-time refinement methods to improve out-of-distribution performance.

Contribution

The paper introduces two novel test-time refinement strategies to mitigate distribution shifts in MLFFs, enhancing their ability to generalize beyond training data without expensive retraining.

Findings

Test-time refinement reduces errors on out-of-distribution systems.

Distribution shifts significantly impact MLFF performance.

Proposed methods are computationally inexpensive and effective.

Abstract

Machine Learning Force Fields (MLFFs) are a promising alternative to expensive ab initio quantum mechanical molecular simulations. Given the diversity of chemical spaces that are of interest and the cost of generating new data, it is important to understand how MLFFs generalize beyond their training distributions. In order to characterize and better understand distribution shifts in MLFFs, we conduct diagnostic experiments on chemical datasets, revealing common shifts that pose significant challenges, even for large foundation models trained on extensive data. Based on these observations, we hypothesize that current supervised training methods inadequately regularize MLFFs, resulting in overfitting and learning poor representations of out-of-distribution systems. We then propose two new methods as initial steps for mitigating distribution shifts for MLFFs. Our methods focus on test-time refinement strategies that incur minimal computational cost and do not use expensive ab initio reference labels. The first strategy, based on spectral graph theory, modifies the edges of test graphs to align with graph structures seen during training. Our second strategy improves representations for out-of-distribution systems at test-time by taking gradient steps using an auxiliary objective, such as a cheap physical prior. Our test-time refinement strategies significantly reduce errors on out-of-distribution systems, suggesting that MLFFs are capable of and can move towards modeling diverse chemical spaces, but are not being effectively trained to do so. Our experiments establish clear benchmarks for evaluating the generalization capabilities of the next generation of MLFFs. Our code is available at https://tkreiman.github.io/projects/mlff_distribution_shifts/.