Role of Structural and Conformational Diversity for Machine Learning Potentials

TL;DR

This paper explores how structural and conformational diversity in datasets affect the generalization of machine learning interatomic potentials, highlighting the importance of balanced data and the limitations of current models.

Contribution

It provides a detailed analysis of the impact of data diversity on MLIP performance and offers guidelines for optimizing quantum mechanics data generation.

Findings

Balanced structural and conformational diversity improves model generalization.

Existing QM datasets often lack the optimal diversity trade-off.

MLIP models struggle to generalize beyond their training distribution.

Abstract

In the field of Machine Learning Interatomic Potentials (MLIPs), understanding the intricate relationship between data biases, specifically conformational and structural diversity, and model generalization is critical in improving the quality of Quantum Mechanics (QM) data generation efforts. We investigate these dynamics through two distinct experiments: a fixed budget one, where the dataset size remains constant, and a fixed molecular set one, which focuses on fixed structural diversity while varying conformational diversity. Our results reveal nuanced patterns in generalization metrics. Notably, for optimal structural and conformational generalization, a careful balance between structural and conformational diversity is required, but existing QM datasets do not meet that trade-off. Additionally, our results highlight the limitation of the MLIP models at generalizing beyond their training distribution, emphasizing the importance of defining applicability domain during model deployment. These findings provide valuable insights and guidelines for QM data generation efforts.