Online Regret Bounds for Markov Decision Processes with Deterministic Transitions.
Ronald OrtnerPublished in: ALT (2008)
Keyphrases
- markov decision processes
- regret bounds
- online learning
- state space
- reinforcement learning
- optimal policy
- finite state
- dynamic programming
- transition matrices
- policy iteration
- infinite horizon
- online convex optimization
- stationary policies
- decision theoretic planning
- partially observable
- action space
- average reward
- learning algorithm
- markov decision process
- linear regression
- reward function
- average cost
- fixed point
- least squares
- computational complexity