On The Finetuning of MLIPs Through the Lens of Iterated Maps With BPTT

TL;DR

This paper introduces a novel fine-tuning method for machine-learning interatomic potentials that directly optimizes predicted structures through an end-to-end differentiable simulation loop, significantly improving accuracy and robustness.

Contribution

It presents a fully-differentiable relaxation process for MLIPs that enhances their predictive accuracy by directly optimizing final structures during fine-tuning.

Findings

Average 32% reduction in prediction error across models

Method is robust to hyperparameter and procedural variations

Improves performance consistently on pretrained MLIPs

Abstract

Accurate structural relaxation is critical for advanced materials design. Traditional approaches built on physics-derived first-principles calculations are computationally expensive, motivating the creation of machine-learning interatomic potentials (MLIPs), which strive to faithfully reproduce first-principles computed forces. We propose a fine-tuning method to be used on a pretrained MLIP in which we create a fully-differentiable end-to-end simulation loop that optimizes the predicted final structures directly. Trajectories are unrolled and gradients are tracked through the entire relaxation. We show that this method consistently improves performance across all evaluated pretrained models; resulting in an average of roughly 32% reduction in prediction error. Interestingly, we show the process is robust to substantial variation in the relaxation setup, achieving negligibly different results across varied hyperparameter and procedural modifications.