BLIPs: Bayesian Learned Interatomic Potentials

TL;DR

BLIPs introduces a Bayesian framework for interatomic potentials that enhances predictive accuracy and provides reliable uncertainty estimates, especially in data-scarce or out-of-distribution scenarios, improving simulation reliability.

Contribution

BLIPs presents a scalable, architecture-agnostic Bayesian approach for training and fine-tuning MLIPs with calibrated uncertainties and minimal computational overhead.

Findings

Improved accuracy over standard MLIPs in chemistry tasks.

Provides well-calibrated uncertainty estimates.

Enhances performance in data-scarce and out-of-distribution regimes.

Abstract

Machine Learning Interatomic Potentials (MLIPs) are becoming a central tool in simulation-based chemistry. However, like most deep learning models, MLIPs struggle to make accurate predictions on out-of-distribution data or when trained in a data-scarce regime, both common scenarios in simulation-based chemistry. Moreover, MLIPs do not provide uncertainty estimates by construction, which are fundamental to guide active learning pipelines and to ensure the accuracy of simulation results compared to quantum calculations. To address this shortcoming, we propose BLIPs: Bayesian Learned Interatomic Potentials. BLIP is a scalable, architecture-agnostic variational Bayesian framework for training or fine-tuning MLIPs, built on an adaptive version of Variational Dropout. BLIP delivers well-calibrated uncertainty estimates and minimal computational overhead for energy and forces prediction at inference time, while integrating seamlessly with (equivariant) message-passing architectures. Empirical results on simulation-based computational chemistry tasks demonstrate improved predictive accuracy with respect to standard MLIPs, and trustworthy uncertainty estimates, especially in data-scarse or heavy out-of-distribution regimes. Moreover, fine-tuning pretrained MLIPs with BLIP yields consistent performance gains and calibrated uncertainties.