Modelling magnetic material properties with uncertainty-aware neural networks

2026-06-10

AI Summary

This work addresses the challenge of unreliable model predictions in magnetic material property forecasting, which arises from data scarcity and out-of-distribution extrapolation. To this end, the authors propose a graph neural network framework integrated with uncertainty quantification. By incorporating Dropout-based Bayesian approximation and a Gaussian negative log-likelihood loss function, the method enables uncertainty-aware modeling for predicting intrinsic properties of permanent magnets and microstructure-driven coercivity. The approach not only significantly enhances the reliability of predictive confidence estimates but also demonstrates, for the first time, the effective transferability of uncertainty quantification strategies across diverse magnetic material modeling tasks. This provides a robust and trustworthy framework for intelligent materials design under limited data conditions.

Abstract

Machine learning is increasingly applied to accelerate the discovery of novel materials by exploring large compositional and structural design spaces. Yet, the scarcity of high-quality data and the frequent need for out-of-distribution prediction introduce substantial uncertainty, making the assessment of model reliability essential. In this work, we investigate uncertainty quantification as a means to evaluate model confidence in the context of permanent magnet research. In a first study, we benchmark classical and modern machine learning models for predicting intrinsic magnetic properties, focusing on the quality of their uncertainty estimates. We apply Gaussian negative log-likelihood loss and dropout-based Bayesian approximation as practical strategies for estimating predictive uncertainty. In a second study, we transfer these architectural features for uncertainty estimation to a more complex task: predicting coercivity from microstructural information using a graph neural network. Together, these studies demonstrate that uncertainty quantification not only enhances the trustworthiness of predictions but is also transferable across different modeling tasks.

Problem

• uncertainty quantification
• magnetic materials
• machine learning
• out-of-distribution prediction
• model reliability

Innovation

• uncertainty quantification
• magnetic materials
• graph neural networks
• Bayesian approximation
• coercivity prediction