Material Property Prediction

Material property prediction aims to accurately forecast a material's characteristics (e.g., band gap, strength) using computational methods, accelerating materials discovery and design. Current research heavily utilizes machine learning, employing graph neural networks (GNNs), large language models (LLMs), and Bayesian neural networks to predict properties from compositional and structural data, with a strong focus on improving model robustness and uncertainty quantification. These advancements are crucial for optimizing material selection in various applications, ranging from energy technologies to biomedical engineering, by reducing the time and cost associated with experimental testing. Furthermore, efforts are underway to improve the handling of out-of-distribution samples and to enhance the interpretability of these predictive models.