An automated approach for developing neural network interatomic potentials with FLAME

TL;DR

This paper introduces an automated, cyclic method for developing neural network interatomic potentials that combines data generation and potential training with minimal human intervention, improving reliability for inorganic systems.

Contribution

It presents a novel automated approach that integrates data generation and potential training in a cyclic process for neural network interatomic potentials.

Findings

Effective in generating diverse training data

Achieves reliable potentials for inorganic systems

Reduces human intervention in potential development

Abstract

The performance of machine learning interatomic potentials relies on the quality of the training dataset. In this work, we present an approach for generating diverse and representative training data points which initiates with \it{ab initio} calculations for bulk structures. The data generation and potential construction further proceed side-by-side in a cyclic process of training the neural network and crystal structure prediction based on the developed interatomic potentials. All steps of the data generation and potential development are performed with minimal human intervention. We show the reliability of our approach by assessing the performance of neural network potentials developed for two inorganic systems.