Out-of-Distribution Material Property Prediction Using Adversarial Learning

The accurate prediction of material properties is crucial in a wide range of scientific and engineering disciplines. Machine learning (ML) has advanced the state of the art in this field, enabling scientists to discover novel materials and design materials with specific desired properties. However, one major challenge that persists in material property prediction is the generalization of models to out-of-distribution (OOD) samples, i.e., samples that differ significantly from those encountered during training. In real-world materials discovery, OOD scenarios often arise when applying ML to predict additional materials within a newly explored region originating from a few experimental samples. In this paper, we explore the application of advancements in OOD learning approaches to enhance the robustness and reliability of material property prediction models. We propose and apply the Crystal Adversarial Learning (CAL) algorithm for OOD materials property prediction, which generates synthetic data during training to guide learning toward those samples with high prediction uncertainty. We further propose an adversarial learning-based targeted approach to make the model adapt to a particular OOD data set, as an alternative to traditional fine-tuning. Our experiments suggest that our CAL algorithm can be effective in ML scenarios with limited samples, which commonly occur in materials science. Our work provides an important step toward improved OOD learning and materials property prediction and highlights areas that require further exploration and refinement.