Performance Improvement Bounds for Lipschitz Configurable Markov Decision Processes.
Alberto Maria MetelliPublished in: CoRR (2024)
Keyphrases
- markov decision processes
- discounted reward
- finite state
- state space
- optimal policy
- dynamic programming
- transition matrices
- factored mdps
- policy iteration
- reinforcement learning
- reachability analysis
- risk sensitive
- decision processes
- upper bound
- lower bound
- infinite horizon
- markov decision process
- upper and lower bounds
- finite horizon
- decision theoretic planning
- reinforcement learning algorithms
- action sets
- model based reinforcement learning
- average reward
- state and action spaces
- partially observable
- reward function
- probabilistic planning
- action space
- average cost
- data mining
- monte carlo
- utility function
- planning under uncertainty