Value Iteration for Long-Run Average Reward in Markov Decision Processes.
Pranav AshokKrishnendu ChatterjeePrzemyslaw DacaJan KretínskýTobias MeggendorferPublished in: CAV (1) (2017)
Keyphrases
- average reward
- long run
- markov decision processes
- optimal policy
- semi markov decision processes
- stochastic games
- policy iteration
- discounted reward
- expected cost
- state space
- dynamic programming
- discount factor
- sample path
- infinite horizon
- average cost
- optimality criterion
- decision problems
- finite horizon
- total reward
- finite state
- reinforcement learning
- state and action spaces
- markov decision process
- reinforcement learning algorithms
- hierarchical reinforcement learning
- control policy
- factored mdps
- multistage
- state dependent
- planning under uncertainty
- sufficient conditions
- queueing networks
- decision processes
- partially observable markov decision processes
- policy gradient
- partially observable
- heuristic search
- reward function