Whittle index based Q-learning for restless bandits with average reward.
Konstantin AvrachenkovVivek S. BorkarPublished in: CoRR (2020)
Keyphrases
- average reward
- optimal policy
- discounted reward
- reinforcement learning
- policy iteration
- model free
- markov decision processes
- state action
- optimal control
- semi markov decision processes
- sample path
- state space
- hierarchical reinforcement learning
- stochastic games
- long run
- optimality criterion
- td learning
- reinforcement learning algorithms
- function approximation
- dynamic programming
- stochastic systems
- actor critic
- rl algorithms
- total reward
- reward function
- semi markov
- state and action spaces
- decision problems
- infinite horizon
- markov decision process
- least squares
- multi agent
- learning algorithm
- reinforcement learning methods
- partially observable
- finite state
- evaluation function
- markov decision problems
- partially observable markov decision processes
- average cost
- search algorithm
- temporal difference
- linear programming
- sufficient conditions
- fixed point