Regret-based Reward Elicitation for Markov Decision Processes
Kevin ReganCraig BoutilierPublished in: CoRR (2012)
Keyphrases
- markov decision processes
- reward function
- minimax regret
- expected reward
- total reward
- reinforcement learning algorithms
- optimal policy
- state space
- dynamic programming
- reinforcement learning
- average reward
- finite state
- finite horizon
- partially observable
- transition matrices
- decision theoretic planning
- markov decision process
- policy iteration
- preference elicitation
- planning under uncertainty
- average cost
- action sets
- factored mdps
- reachability analysis
- model based reinforcement learning
- utility elicitation
- risk sensitive
- action space
- state and action spaces
- decision processes
- infinite horizon
- stationary policies
- learning agent
- utility function
- multistage
- learning algorithm