Function approximation for reinforcement learning based on reaction-diffusion equation on a graph | IEEE Conference Publication | IEEE Xplore