A generalizable, uncertainty-aware neural network potential for GeSbTe with Monte Carlo dropout. (January 2023)

• Record Type: Journal Article
• Authors: Lee, Sung-Ho; Olevano, Valerio; Sklénard, Benoit
• Is Part Of: Solid-state electronics. Volume 199(2023)
• Publication Date: 2023-01
• Subjects: Neural network potential -- Machine learning -- Atomistic simulations -- MD simulation -- GST; Semiconductors -- Periodicals; Semiconducteurs -- Périodiques; 621.38152
• DOI: 10.1016/j.sse.2022.108508
• Languages: English
• ISSNs: 0038-1101

Abstract

Abstract: A Bayesian neural network potential (NNP) achieved with the Monte Carlo dropout approximation method is developed for GeSbTe alloys. The Bayesian NNP is shown to be more generalizable than its classical counterpart, yielding reasonable predictions on structures that are not directly in the training configurations, and is able to output uncertainty estimates for the predictions. Its application to a molecular dynamics (MD) simulation is also presented, and the validity of the obtained trajectory is evaluated by comparing it to Density Functional Theory (DFT). Highlights: Neural network potentials can run simulations at previously impossible scales. Bayesian approach can be used to output the predictive uncertainties. They also show better generalizability than classical neural network potentials.