Ergodic Unobservable MDPs: Decidability of Approximation.
Krishnendu ChatterjeeDavid LurieRaimundo SaonaBruno ZiliottoPublished in: CoRR (2024)
Keyphrases
- markov decision processes
- state space
- factored mdps
- markov chain
- error bounds
- reinforcement learning
- efficient computation
- description logics
- queueing networks
- closed form
- optimal policy
- approximation algorithms
- partially observable
- planning under uncertainty
- dynamic programming
- reward function
- initial state
- stationary distribution
- markov process
- approximation error
- approximation methods