Comparing Policies in Markov Decision Processes: Mandl's Lemma Revisited.
Adam ShwartzArmand M. MakowskiPublished in: Math. Oper. Res. (1990)
Keyphrases
- markov decision processes
- optimal policy
- markov decision process
- decision processes
- average cost
- reinforcement learning
- reward function
- state space
- finite state
- decentralized control
- finite horizon
- partially observable markov decision processes
- policy iteration algorithm
- total reward
- dynamic programming
- discounted reward
- infinite horizon
- macro actions
- transition matrices
- action space
- decision problems
- average reward
- policy iteration
- reachability analysis
- reinforcement learning algorithms
- decision theoretic planning
- multistage
- control policies
- partially observable
- markov decision problems
- long run
- planning under uncertainty
- model based reinforcement learning
- factored mdps
- state and action spaces
- expected reward
- semi markov decision processes
- sufficient conditions
- stationary policies
- risk sensitive
- machine learning
- action sets
- hierarchical reinforcement learning
- probabilistic planning