A perturbation approach to approximate value iteration for average cost Markov decision processes with Borel spaces and bounded costs.
Oscar Vega-AmayaJoaquín López-BorbónPublished in: Kybernetika (2019)
Keyphrases
- markov decision processes
- average cost
- finite state
- policy iteration
- optimal policy
- finite horizon
- reinforcement learning
- temporal difference learning
- stationary policies
- decision theoretic planning
- state space
- infinite horizon
- planning under uncertainty
- markov decision process
- initial state
- reinforcement learning algorithms
- dynamic programming
- risk sensitive
- action sets
- decision processes
- average reward
- partially observable
- reward function
- action space
- learning algorithm
- markov decision problems