Follow-the-Perturbed-Leader for Adversarial Markov Decision Processes with Bandit Feedback.
Yan DaiHaipeng LuoLiyu ChenPublished in: CoRR (2022)
Keyphrases
- markov decision processes
- optimal policy
- finite state
- state space
- reinforcement learning
- policy iteration
- dynamic programming
- decision theoretic planning
- transition matrices
- risk sensitive
- average reward
- finite horizon
- reachability analysis
- decision processes
- reinforcement learning algorithms
- planning under uncertainty
- partially observable
- multi agent
- infinite horizon
- model based reinforcement learning
- average cost
- decision making
- action space
- reward function
- factored mdps
- action sets
- original data
- real time dynamic programming
- semi markov decision processes
- discounted reward
- state and action spaces
- decision diagrams
- markov decision process