Exploiting structure and uncertainty of Bellman updates in Markov decision processes.
Davide TateoCarlo D'EramoAlessandro NuaraMarcello RestelliAndrea BonariniPublished in: SSCI (2017)
Keyphrases
- markov decision processes
- optimal policy
- finite state
- state space
- policy iteration
- dynamic programming
- reinforcement learning algorithms
- reachability analysis
- infinite horizon
- decision processes
- reinforcement learning
- decision theoretic planning
- markov decision process
- transition matrices
- finite horizon
- factored mdps
- risk sensitive
- average reward
- average cost
- planning under uncertainty
- partially observable
- semi markov decision processes
- action space
- model based reinforcement learning
- sufficient conditions
- interval estimation
- stochastic shortest path