Tight Regret Bounds for Model-Based Reinforcement
      Learning with Greedy Policies

  Yonathan Efroni     Nadav Merlis     Mohammad Ghavamzadeh        Shie Mannor
  Technion, Israel     Technion, Israel    Facebook AI Research       Technion, Israel

                         Abstract
       State-of-the-art efficient model-based Reinforcement Learning (RL) algorithms typ-
       ically act by iteratively solving empirical models, i.e., by performing full-planning
       on Markov Decision P ...