# Dynamic Training Enhances Machine Learning Potentials for Long-Lasting Molecular Dynamics

## TL;DR

Dynamic training improves machine learning models for long-term molecular simulations, offering better accuracy and broader applicability.

## Contribution

Dynamic training is introduced as an architecture-independent method to enhance accuracy in long-lasting molecular dynamics simulations.

## Key findings

- Dynamic training applied to an EGNN improved prediction accuracy for a hydrogen-palladium system on graphene.
- The method is architecture-independent, making it applicable to various machine learning potentials.
- DT offers a practical tool for advancing molecular dynamics simulations with enhanced accuracy.

## Abstract

Molecular dynamics
(MD) simulations are vital for exploring
complex
systems in computational physics and chemistry. While machine learning
methods dramatically reduce computational costs relative to ab initio
methods, their accuracy in long-lasting simulations remains limited.
Here we propose dynamic training (DT), a method designed to enhance
accuracy of a model over extended MD simulations. Applying DT to an
equivariant graph neural network (EGNN) on the challenging system
of a hydrogen molecule interacting with a palladium cluster anchored
to a graphene vacancy demonstrates a superior prediction accuracy
compared to conventional approaches. Crucially, the DT architecture-independent
design ensures its applicability across diverse machine learning potentials,
making it a practical tool for advancing MD simulations.