Uncertainty-biased molecular dynamics for learning uniformly accurate interatomic potentials - Archive ouverte HAL
Article Dans Une Revue npj Computational Materials Année : 2024

Uncertainty-biased molecular dynamics for learning uniformly accurate interatomic potentials

Résumé

Efficiently creating a concise but comprehensive data set for training machine-learned interatomic potentials (MLIPs) is an under-explored problem. Active learning, which uses biased or unbiased molecular dynamics (MD) to generate candidate pools, aims to address this objective. Existing biased and unbiased MD-simulation methods, however, are prone to miss either rare events or extrapolative regions—areas of the configurational space where unreliable predictions are made. This work demonstrates that MD, when biased by the MLIP’s energy uncertainty, simultaneously captures extrapolative regions and rare events, which is crucial for developing uniformly accurate MLIPs. Furthermore, exploiting automatic differentiation, we enhance bias-forces-driven MD with the concept of bias stress. We employ calibrated gradient-based uncertainties to yield MLIPs with similar or, sometimes, better accuracy than ensemble-based methods at a lower computational cost. Finally, we apply uncertainty-biased MD to alanine dipeptide and MIL-53(Al), generating MLIPs that represent both configurational spaces more accurately than models trained with conventional MD.
Fichier principal
Vignette du fichier
Uncertainty-biased molecular dynamics for learning uniformly accurate interatomic potentials (npjCompMat).pdf (7.57 Mo) Télécharger le fichier
Origine Publication financée par une institution
Licence

Dates et versions

hal-04712228 , version 1 (27-09-2024)

Licence

Identifiants

Citer

Viktor Zaverkin, David Holzmüller, Henrik Christiansen, Federico Errica, Francesco Alesiani, et al.. Uncertainty-biased molecular dynamics for learning uniformly accurate interatomic potentials. npj Computational Materials, 2024, npj Computational Materials, 10 (1), pp.83. ⟨10.1038/s41524-024-01254-1⟩. ⟨hal-04712228⟩
0 Consultations
0 Téléchargements

Altmetric

Partager

More