Unichain and Aperiodicity are Sufficient for Asymptotic Optimality of Average-Reward Restless Bandits.
Yige HongQiaomin XieYudong ChenWeina WangPublished in: CoRR (2024)
Keyphrases
- average reward
- asymptotic optimality
- optimal policy
- finite state
- asymptotically optimal
- markov decision processes
- initial state
- average cost
- sufficient conditions
- long run
- policy iteration
- dynamic programming
- state space
- reinforcement learning
- decision problems
- partially observable markov decision processes
- optimal control
- infinite horizon
- stationary policies
- discounted reward
- markov chain
- multistage
- markov decision process
- model free
- flowshop
- stochastic systems
- evolutionary algorithm
- markov decision problems
- lower bound
- learning algorithm
- neural network