Frozen Policy Iteration: Computationally Efficient RL under Linear $Q^{\ddot{I}}€$ Realizability for Deterministic Dynamics

Yijing Ke, Zihan Zhang, Ruosong Wang

We study computationally and statistically efficient reinforcement learning under the linear $Q^{\ddot{I}}€$ realizability assumption, where any policy's $Q$-function is linear in a given s…