Learning atomic forces from uncertainty-calibrated adversarial attacks

TL;DR

This paper introduces CAGO, a novel algorithm that uses uncertainty calibration and adversarial structure generation to improve machine learning interatomic potentials, reducing training data needs and enhancing prediction reliability.

Contribution

The paper presents CAGO, a method for generating adversarial structures with controlled errors, improving MLIP training efficiency and robustness through uncertainty calibration.

Findings

CAGO effectively generates adversarial structures with user-defined errors.

MLIPs trained with CAGO converge faster, requiring fewer training structures.

The approach enhances the stability and accuracy of MLIPs for complex systems.

Abstract

Adversarial approaches, which intentionally challenge machine learning models by generating difficult examples, are increasingly being adopted to improve machine learning interatomic potentials (MLIPs). While already providing great practical value, little is known about the actual prediction errors of MLIPs on adversarial structures and whether these errors can be controlled. We propose the Calibrated Adversarial Geometry Optimization (CAGO) algorithm to discover adversarial structures with user-assigned errors. Through uncertainty calibration, the estimated uncertainty of MLIPs is unified with real errors. By performing geometry optimization for calibrated uncertainty, we reach adversarial structures with the user-assigned target MLIP prediction error. Integrating with active learning pipelines, we benchmark CAGO, demonstrating stable MLIPs that systematically converge structural, dynamical, and thermodynamical properties for liquid water and water adsorption in a metal-organic framework within only hundreds of training structures, where previously many thousands were typically required.