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

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, simultaneous training NNs on energies forces, which a prerequisite for, e.g., molecular dynamics simulations, can be demanding. In this work, we present an improved NN architecture based previous GM-NN model [V. Zaverkin J. K\"astner, Chem. Theory Comput. 16, 5410-5421 (2020)], shows prediction accuracy considerably reduced times. Moreover, extend applicability Gaussian moment-based periodic systems demonstrate overall excellent transferability robustness respective models. The fast by methodology is pre-requisite for training-heavy workflows such as active or learning-on-the-fly.