EUVIMEDEUVIMED entdecken
Crossref

Scalable Bayesian Uncertainty Quantification for Neural Network Potentials: Promise and Pitfalls

Stephan Thaler iD, Gregor Doehner iD, Julija Zavadlav iD

Journal of Chemical Theory and Computation · 25.07.2023 · 19 (14) · 4520-4532

DOI: 10.1021/acs.jctc.2c01267 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Chemical Theory and Computation, online erschienen am 04.04.2023. Verfasst von Stephan Thaler und 2 weiteren Autor:innen. Beteiligte Einrichtung: Professorship of Multiscale Modeling of Fluid Materials, Department of Engineering Physics and Computation, TUM School of Engineering and Design, Tec…. Die Arbeit wurde bisher 35-mal zitiert (Crossref) und nennt 89 Literaturangaben. Gefördert von: Munich Data Science Institute, Technical University of Munich.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Neural network (NN) potentials promise highly accurate molecular dynamics (MD) simulations within the computational complexity of classical MD force fields. However, when applied outside their training domain, NN potential predictions can be inaccurate, increasing the need for Uncertainty Quantification (UQ). Bayesian modeling provides the mathematical framework for UQ, but classical Bayesian methods based on Markov chain Monte Carlo (MCMC) are computationally intractable for NN potentials. By training graph NN potentials for coarse-grained systems of liquid water and alanine dipeptide, we demonstrate here that scalable Bayesian UQ via stochastic gradient MCMC (SG-MCMC) yields reliable uncertainty estimates for MD observables. We show that cold posteriors can reduce the required training data size and that for reliable UQ, multiple Markov chains are needed. Additionally, we find that SG-MCMC and the Deep Ensemble method achieve comparable results, despite shorter training and less hyperparameter tuning of the latter. We show that both methods can capture aleatoric and epistemic uncertainty reliably, but not systematic uncertainty, which needs to be minimized by adequate modeling to obtain accurate credible intervals for MD observables. Our results represent a step toward accurate UQ that is of vital importance for trustworthy NN potential-based MD simulations required for decision-making in practice.

Autor:innen und Institutionen

  • Professorship of Multiscale Modeling of Fluid Materials, Department of Engineering Physics and Computation, TUM School of Engineering and Design, Technical University of Munich, 85748 Garching near Munich, Germany.
  • Munich Data Science Institute, Technical University of Munich, 85748 Garching near Munich, Germany.

ORCID-Profile: Stephan Thaler, Gregor Doehner, Julija Zavadlav

Förderung

  • Munich Data Science Institute, Technical University of Munich

Zitieren

APA

Thaler, S., Doehner, G., & Zavadlav, J. (2023). Scalable Bayesian Uncertainty Quantification for Neural Network Potentials: Promise and Pitfalls. Journal of Chemical Theory and Computation, 19(14), 4520-4532. https://doi.org/10.1021/acs.jctc.2c01267

Vancouver

Thaler S, Doehner G, Zavadlav J. Scalable Bayesian Uncertainty Quantification for Neural Network Potentials: Promise and Pitfalls. Journal of Chemical Theory and Computation. 2023;19(14):4520-4532. doi:10.1021/acs.jctc.2c01267

BibTeX
@article{Thaler2023,
  author = {Thaler, Stephan and Doehner, Gregor and Zavadlav, Julija},
  title = {Scalable Bayesian Uncertainty Quantification for Neural Network Potentials: Promise and Pitfalls},
  journal = {Journal of Chemical Theory and Computation},
  year = {2023},
  volume = {19},
  number = {14},
  pages = {4520--4532},
  doi = {10.1021/acs.jctc.2c01267}
}
Kostenlos starten →