Re-understanding Finite-State Representations of Recurrent Policy Networks.
Mohamad H. DaneshAnurag KoulAlan FernSaeed KhorramPublished in: ICML (2021)
Keyphrases
- finite state
- optimal policy
- markov chain
- partially observable markov decision processes
- markov decision processes
- policy iteration algorithm
- average cost
- policy iteration
- model checking
- continuous state
- state space
- decision problems
- action sets
- context free
- dynamic programming
- tree automata
- average reward
- state dependent
- continuous time bayesian networks
- reinforcement learning
- reward function
- vector quantizer
- action selection
- infinite horizon
- sufficient conditions
- finite state transducers
- probabilistic model
- search algorithm
- data mining