Optimistic Policy Iteration for MDPs with Acyclic Transient State Structure.
Joseph LubarsAnna WinnickiMichael LivesayR. SrikantPublished in: CoRR (2021)
Keyphrases
- policy iteration
- markov decision processes
- state space
- markov decision process
- optimal policy
- factored mdps
- reinforcement learning
- discounted reward
- fixed point
- model free
- sample path
- average reward
- policy evaluation
- markov decision problems
- finite state
- temporal difference
- approximate dynamic programming
- action space
- transition matrices
- partially observable
- state variables
- linear programming
- least squares
- average cost
- np hard
- dynamic programming
- partially observable markov decision processes
- context specific
- convergence rate
- particle filter
- search space