Active learning of linearly parametrized interatomic potentials

[Display omitted] This paper introduces an active learning approach to the fitting of machine learning interatomic potentials. Our approach is based on the D-optimality criterion for selecting atomic configurations on which the potential is fitted. It is shown that the proposed active learning appro...

Full description

Saved in:
Bibliographic Details
Published inComputational materials science Vol. 140; pp. 171 - 180
Main Authors Podryabinkin, Evgeny V., Shapeev, Alexander V.
Format Journal Article
LanguageEnglish
Published Elsevier B.V 01.12.2017
Subjects
Active learning
Atomistic simulation
Interatomic potential
Learning on the fly
Machine learning
Moment tensor potentials
Atomistic simulation
Interatomic potential
Active learning
Learning on the fly
Moment tensor potentials
Machine learning
Online AccessGet full text

Cover

More Information
Summary:[Display omitted] This paper introduces an active learning approach to the fitting of machine learning interatomic potentials. Our approach is based on the D-optimality criterion for selecting atomic configurations on which the potential is fitted. It is shown that the proposed active learning approach is highly efficient in training potentials on the fly, ensuring that no extrapolation is attempted and leading to a completely reliable atomistic simulation without any significant decrease in accuracy. We apply our approach to molecular dynamics and structure relaxation, and we argue that it can be applied, in principle, to any other type of atomistic simulation. The software, test cases, and examples of usage are published at http://gitlab.skoltech.ru/shapeev/mlip/.
ISSN:0927-0256
1879-0801
DOI:10.1016/j.commatsci.2017.08.031