Constrained Markov Decision Processes via Backward Value Functions.
Harsh SatijaPhilip AmortilaJoelle PineauPublished in: ICML (2020)
Keyphrases
- markov decision processes
- finite state
- state space
- optimal policy
- transition matrices
- dynamic programming
- reinforcement learning
- policy iteration
- risk sensitive
- reachability analysis
- decision theoretic planning
- infinite horizon
- decision diagrams
- planning under uncertainty
- decision processes
- finite horizon
- average cost
- reinforcement learning algorithms
- action space
- markov decision process
- action sets
- model based reinforcement learning
- average reward
- factored mdps
- state and action spaces
- partially observable
- machine learning
- search algorithm
- multi valued
- discounted reward
- semi markov decision processes