We use Laplacian representation to improve exploration for reinforcement learning agents.