Crystal-graph attention networks for the prediction of stable materials

Graph neural networks have enjoyed great success in the prediction of material properties for both molecules and crystals. These networks typically use the atomic positions...