Measuring transferability issues in machine-learning force fields: The example of Gold-Iron interactions with linearized potentials

TL;DR

This paper investigates the transferability challenges of machine-learning force fields, specifically for gold-iron interactions, highlighting the trade-off between model complexity and generalization in nanoparticle simulations.

Contribution

It introduces a linearized potential with penalizing regression to analyze transferability issues and demonstrates the impact of model complexity on accuracy and overfitting.

Findings

More complex models fit training data better

Higher complexity can cause overfitting and reduce accuracy on new systems

Linearized potentials help control model complexity

Abstract

Machine-learning force fields have been increasingly employed in order to extend the possibility of current first-principles calculations. However, the transferability of the obtained potential can not always be guaranteed in situations that are outside the original database. To study such limitation, we examined the very difficult case of the interactions in gold-iron nanoparticles. For the machine-learning potential, we employed a linearized formulation that is parameterized using a penalizing regression scheme which allows us to control the complexity of the obtained potential. We showed that while having a more complex potential allows for a better agreement with the training database, it can also lead to overfitting issues and a lower accuracy in untrained systems.