Fast and Sample-Efficient Interatomic Neural Network Potentials for Molecules and Materials Based on Gaussian Moments

TL;DR

This paper introduces an improved neural network architecture based on Gaussian moments for interatomic potentials, achieving faster training and better accuracy, applicable to molecules and periodic materials, facilitating large-scale simulations.

Contribution

An enhanced GM-NN model that improves prediction accuracy, reduces training time, and extends applicability to periodic systems for more efficient atomistic simulations.

Findings

Faster training times compared to previous models

High transferability to periodic systems

Robust performance across diverse molecular and material datasets

Abstract

Artificial neural networks (NNs) are one of the most frequently used machine learning approaches to construct interatomic potentials and enable efficient large-scale atomistic simulations with almost ab initio accuracy. However, the simultaneous training of NNs on energies and forces, which are a prerequisite for, e.g., molecular dynamics simulations, can be demanding. In this work, we present an improved NN architecture based on the previous GM-NN model [V. Zaverkin and J. K\"astner, J. Chem. Theory Comput. 16, 5410-5421 (2020)], which shows an improved prediction accuracy and considerably reduced training times. Moreover, we extend the applicability of Gaussian moment-based interatomic potentials to periodic systems and demonstrate the overall excellent transferability and robustness of the respective models. The fast training by the improved methodology is a pre-requisite for training-heavy workflows such as active learning or learning-on-the-fly.