Pessimistic Q-Learning for Offline Reinforcement Learning: Towards Optimal Sample Complexity.
Laixi ShiGen LiYuting WeiYuxin ChenYuejie ChiPublished in: CoRR (2022)
Keyphrases
- reinforcement learning
- sample complexity
- sequential decision problems
- learning algorithm
- learning problems
- supervised learning
- function approximation
- model free
- dynamic programming
- optimal control
- state space
- reinforcement learning algorithms
- optimal policy
- pac learning
- active exploration
- multi agent
- theoretical analysis
- active learning
- temporal difference
- upper bound
- machine learning
- vc dimension
- markov decision processes
- lower bound
- generalization error
- multi agent reinforcement learning
- action selection
- optimal solution
- training data
- worst case
- special case
- small number
- decision problems
- transfer learning
- machine learning algorithms
- average case
- semi supervised learning
- partially observable
- control policy
- continuous state and action spaces
- high dimensional