Pessimistic Q-Learning for Offline Reinforcement Learning: Towards Optimal Sample Complexity

Open in new window