Feature selection in reinforcement learning (RL), i.e. choosing basis functions such that useful approximations of the unknown value function can be obtained, is one of the main challenges in scaling ...