PAC Bounds for Imitation and Model-based Batch Learning of Contextual Markov Decision Processes.
Yash NairFinale Doshi-VelezPublished in: CoRR (2020)
Keyphrases
- markov decision processes
- batch learning
- reinforcement learning
- upper bound
- vc dimension
- incremental learning
- state space
- discounted reward
- optimal policy
- finite state
- interval estimation
- lower bound
- policy iteration
- model free
- transition matrices
- sample complexity
- dynamic programming
- average reward
- upper and lower bounds
- concept drift
- infinite horizon
- finite horizon
- action space
- reachability analysis
- model based reinforcement learning
- planning under uncertainty
- average cost
- action sets
- partially observable
- decision theoretic planning
- state and action spaces
- reinforcement learning algorithms
- markov decision process
- sample size
- linear programming
- supervised learning