Convergence for Natural Policy Gradient on Infinite-State Average-Reward Markov Decision Processes.
Isaac GrosofSiva Theja MaguluriR. SrikantPublished in: CoRR (2024)
Keyphrases
- average reward
- markov decision processes
- policy gradient
- policy gradient reinforcement learning
- optimal policy
- actor critic
- policy iteration
- state space
- finite state
- semi markov decision processes
- reinforcement learning
- stochastic games
- state action
- long run
- dynamic programming
- discounted reward
- reinforcement learning algorithms
- convergence rate
- average cost
- total reward
- state and action spaces
- model free
- partially observable
- markov chain
- decision problems
- stationary policies
- convergence speed
- action space
- markov decision process
- reward function
- infinite horizon
- dynamic environments
- reinforcement learning methods
- machine learning