Neural-network Density Functional Theory Based on Variational Energy Minimization

Instead of training the neural networks with predictions derived from DFT, unsupervised learning by treating the energy functional of variational DFT as the loss function of the neural network yields predictions of material properties.