Online Markov Decision Processes with Aggregate Bandit Feedback.
Alon CohenHaim KaplanTomer KorenYishay MansourPublished in: COLT (2021)
Keyphrases
- markov decision processes
- finite state
- state space
- reinforcement learning
- planning under uncertainty
- optimal policy
- reachability analysis
- partially observable
- transition matrices
- online learning
- reinforcement learning algorithms
- dynamic programming
- factored mdps
- average cost
- decision theoretic planning
- policy iteration
- infinite horizon
- average reward
- action sets
- finite horizon
- markov decision process
- reward function
- model based reinforcement learning
- decision diagrams
- risk sensitive
- state abstraction
- total reward
- multistage
- real time dynamic programming