Adaptive Bases for Reinforcement Learning

Dotan Di Castro, Shie Mannor

We consider the problem of reinforcement learning using function approximation, where the approximating basis can change dynamically while interacting with the environment. A motiv…