Twice Regularized Markov Decision Processes: The Equivalence between Robustness and Regularization.
Esther DermanYevgeniy MenMatthieu GeistShie MannorPublished in: CoRR (2023)
Keyphrases
- markov decision processes
- regularization method
- state space
- optimal policy
- policy iteration
- finite state
- transition matrices
- dynamic programming
- decision diagrams
- reinforcement learning
- planning under uncertainty
- factored mdps
- reinforcement learning algorithms
- reachability analysis
- finite horizon
- average reward
- partially observable
- average cost
- markov decision process
- action space
- decision theoretic planning
- infinite horizon
- objective function
- decision processes
- reward function
- regularization parameter
- convergence rate
- fixed point
- decision problems
- image restoration
- least squares
- real time dynamic programming
- interval estimation