Improved Sample Complexity Analysis of Natural Policy Gradient Algorithm with General Parameterization for Infinite Horizon Discounted Reward Markov Decision Processes.
Washim Uddin MondalVaneet AggarwalPublished in: AISTATS (2024)
Keyphrases
- markov decision processes
- average reward
- policy iteration
- infinite horizon
- complexity analysis
- optimal policy
- dynamic programming
- discounted reward
- policy gradient
- long run
- finite horizon
- actor critic
- reinforcement learning
- reinforcement learning algorithms
- state space
- average cost
- partially observable
- finite state
- model free
- optimality criterion
- state action
- learning algorithm
- markov decision process
- action space
- stochastic games
- least squares
- optimal solution
- partially observable markov decision processes
- optimal control
- cost function
- decision making
- monte carlo
- computational complexity
- mathematical model