Robustness and risk-sensitivity in Markov decision processes.
Takayuki OsogamiPublished in: NIPS (2012)
Keyphrases
- markov decision processes
- risk sensitive
- optimal policy
- state space
- policy iteration
- finite state
- dynamic programming
- reinforcement learning
- reinforcement learning algorithms
- transition matrices
- finite horizon
- infinite horizon
- factored mdps
- decision theoretic planning
- partially observable
- reachability analysis
- state and action spaces
- action space
- planning under uncertainty
- average reward
- decision making
- decision processes
- sensitivity analysis
- markov decision process
- model based reinforcement learning
- average cost
- action sets
- machine learning
- semi markov decision processes
- state abstraction
- decision diagrams
- utility function
- partially observable markov decision processes