Interpretable Ensemble Learning for Materials Property Prediction with Classical Interatomic Potentials: Carbon as an Example

TL;DR

This paper introduces an interpretable ensemble learning method using regression trees that predicts materials properties directly from classical interatomic potentials, avoiding complex descriptors and improving accuracy for carbon allotropes.

Contribution

It presents a novel ensemble learning approach that predicts materials properties directly from interatomic potentials, enhancing interpretability and accuracy without requiring descriptors.

Findings

Ensemble learning outperforms classical interatomic potentials in accuracy.

The method predicts formation energy and elastic constants effectively.

It operates without the need for complex structural descriptors.

Abstract

Machine learning (ML) is widely used to explore crystal materials and predict their properties. However, the training is time-consuming for deep-learning models, and the regression process is a black box that is hard to interpret. Also, the preprocess to transfer a crystal structure into the input of ML, called descriptor, needs to be designed carefully. To efficiently predict important properties of materials, we propose an approach based on ensemble learning consisting of regression trees to predict formation energy and elastic constants based on small-size datasets of carbon allotropes as an example. Without using any descriptor, the inputs are the properties calculated by molecular dynamics with 9 different classical interatomic potentials. Overall, the results from ensemble learning are more accurate than those from classical interatomic potentials, and ensemble learning can capture the relatively accurate properties from the 9 classical potentials as criteria for predicting the final properties.