Title: Learning DFT

Abstract: We present an extension of reverse engineered Kohn-Sham potentials from a density matrix renormalization group calculation towards the construction of a density functional theory functional via deep learning. Instead of applying machine learning to the energy functional itself, we apply these techniques to the Kohn-Sham potentials. To this end we develop a scheme to train a neural network to represent the mapping from local densities to Kohn-Sham potentials. Finally, we use the neural network to up-scale the simulation to larger system sizes.