Regret Analysis of Policy Gradient Algorithm for Infinite Horizon Average Reward Markov Decision Processes.
Qinbo BaiWashim Uddin MondalVaneet AggarwalPublished in: CoRR (2023)
Keyphrases
- average reward
- markov decision processes
- total reward
- optimal policy
- infinite horizon
- long run
- policy iteration
- policy gradient
- dynamic programming
- finite horizon
- reinforcement learning
- actor critic
- state space
- stochastic games
- model free
- finite state
- reinforcement learning algorithms
- discount factor
- partially observable
- average cost
- reward function
- discounted reward
- action space
- partially observable markov decision processes
- least squares
- objective function
- initial state
- markov decision process
- optimal control
- convergence rate
- linear programming