The MLIP package: Moment Tensor Potentials with MPI and Active Learning

TL;DR

This paper details the construction of moment tensor potentials using active learning within the MLIP package, emphasizing efficient sampling, training set expansion, and cost-effective ab initio calculations for interatomic potentials.

Contribution

It introduces methods for constructing moment tensor potentials with active learning, focusing on efficient sampling and training strategies in the MLIP package.

Findings

Active learning improves sampling efficiency for training potentials.

Expanding the training set reduces prediction errors.

Cost-effective setup of ab initio calculations enhances potential accuracy.

Abstract

The subject of this paper is the technology (the "how") of constructing machine-learning interatomic potentials, rather than science (the "what" and "why") of atomistic simulations using machine-learning potentials. Namely, we illustrate how to construct moment tensor potentials using active learning as implemented in the MLIP package, focusing on the efficient ways to sample configurations for the training set, how expanding the training set changes the error of predictions, how to set up ab initio calculations in a cost-effective manner, etc. The MLIP package (short for Machine-Learning Interatomic Potentials) is available at https://mlip.skoltech.ru/download/.