Deep learning, which has achieved significant success in image classification, was quickly introduced into materials research and development. The trigger for this advancement was the Crystal Graph Convolutional Neural Networks (CGCNN) model.
By leveraging graph theory to efficiently extract features from structural information of materials and combining it with deep learning, CGCNN drastically improved the accuracy of material property predictions. Since then, various enhancements have been made, and active research in this area continues to this day.
After CGCNN, numerous models based on graph theory have been reported. For more details, please refer to the following review paper.
Graph neural networks for materials science and chemistry (2022)
Graph neural networks: A review of methods and applications (2020)
Benchmarking graph neural networks for materials chemistry (2021)
A review on the applications of graph neural networks in materials science at the atomic scale (2024)
A review on the applications of graph neural networks in materials science at the atomic scale