Model-Free, Regret-Optimal Best Policy Identification in Online CMDPs.
Zihan ZhouHonghao WeiLei YingPublished in: CoRR (2023)
Keyphrases
- model free
- total reward
- average reward
- reinforcement learning algorithms
- reinforcement learning
- online learning
- policy iteration
- temporal difference
- function approximation
- worst case
- optimal policy
- markov decision processes
- policy evaluation
- control policy
- online algorithms
- dynamic programming
- lower bound
- long run
- pattern recognition
- neural network
- online convex optimization
- regret bounds
- finite horizon
- state space
- optimal solution
- training data