Fast and Fourier Features for Transfer Learning of Interatomic Potentials

TL;DR

This paper introduces Franken, a transfer learning framework using Fourier features and pretrained graph neural networks, enabling fast, accurate, and data-efficient training of interatomic potentials for diverse systems.

Contribution

Franken is a novel transfer learning method that leverages Fourier features and pretrained GNNs to efficiently adapt interatomic potentials without hyperparameter tuning.

Findings

Outperforms kernel-based methods in speed and accuracy on transition metals

Reduces training time from hours to minutes on a single GPU

Achieves stable potentials for water and interfaces with minimal data

Abstract

Training machine learning interatomic potentials that are both computationally and data-efficient is a key challenge for enabling their routine use in atomistic simulations. To this effect, we introduce franken, a scalable and lightweight transfer learning framework that extracts atomic descriptors from pretrained graph neural networks and transfer them to new systems using random Fourier features-an efficient and scalable approximation of kernel methods. Franken enables fast and accurate adaptation of general-purpose potentials to new systems or levels of quantum mechanical theory without requiring hyperparameter tuning or architectural modifications. On a benchmark dataset of 27 transition metals, franken outperforms optimized kernel-based methods in both training time and accuracy, reducing model training from tens of hours to minutes on a single GPU. We further demonstrate the framework's strong data-efficiency by training stable and accurate potentials for bulk water and the Pt(111)/water interface using just tens of training structures. Our open-source implementation (https://franken.readthedocs.io) offers a fast and practical solution for training potentials and deploying them for molecular dynamics simulations across diverse systems.