Concentration bounds for SSP Q-learning for average cost MDPs.
Shaan Ul HaqueVivek S. BorkarPublished in: CoRR (2022)
Keyphrases
- average cost
- markov decision processes
- optimal policy
- discounted reward
- policy iteration
- reinforcement learning
- state space
- reinforcement learning algorithms
- long run
- finite state
- finite horizon
- average reward
- infinite horizon
- finite number
- initial state
- stochastic shortest path
- approximate dynamic programming
- asymptotically optimal
- markov decision process
- markov decision chains
- multistage
- discount factor
- markov decision problems
- decision problems
- function approximation
- linear program
- risk sensitive
- action sets
- optimal control
- dynamic programming
- inventory models
- lower bound
- temporal difference learning
- reward function
- partially observable
- multi agent
- stationary policies
- machine learning
- sufficient conditions
- total cost
- real time
- model free
- control policies
- action selection
- continuous state
- inventory level
- state dependent
- markov chain
- linear programming
- state and action spaces
- optimal solution
- action space
- decision making
- learning algorithm