Online learning in Markov decision processes with arbitrarily changing rewards and transitions.
Jia Yuan YuShie MannorPublished in: GAMENETS (2009)
Keyphrases
- markov decision processes
- online learning
- reinforcement learning
- state space
- optimal policy
- finite state
- transition matrices
- dynamic programming
- decision theoretic planning
- partially observable
- finite horizon
- planning under uncertainty
- policy iteration
- factored mdps
- e learning
- risk sensitive
- reward function
- reinforcement learning algorithms
- infinite horizon
- active learning
- average cost
- decision processes
- sequential decision making under uncertainty
- decentralized control
- markov decision process
- action sets
- reachability analysis
- state transition
- interval estimation