Optimality and Approximation with Policy Gradient Methods in Markov Decision Processes.
Alekh AgarwalSham M. KakadeJason D. LeeGaurav MahajanPublished in: COLT (2020)
Keyphrases
- markov decision processes
- average reward
- average cost
- policy gradient methods
- policy gradient
- stationary policies
- dynamic programming
- actor critic
- finite state
- optimal policy
- state space
- reinforcement learning
- policy iteration
- planning under uncertainty
- natural actor critic
- markov decision process
- reinforcement learning algorithms
- action space
- partially observable
- approximate dynamic programming
- optimal solution
- infinite horizon
- decision processes
- approximation methods
- dynamic environments
- stochastic games
- partially observable markov decision processes
- long run
- decision problems