Reinforcement Learning with Logarithmic Regret and Policy Switches

In this paper, we study the problem of regret minimization for episodic Reinforcement Learning (RL) both in the model-free and the model-based setting.