Markov Abstractions for PAC Reinforcement Learning in Non-Markov Decision Processes.
Alessandro RoncaGabriel Paludo LicksGiuseppe De GiacomoPublished in: IJCAI (2022)
Keyphrases
- markov decision processes
- reinforcement learning
- reinforcement learning algorithms
- state space
- optimal policy
- finite state
- markov chain
- action space
- function approximation
- policy iteration
- state and action spaces
- dynamic programming
- model based reinforcement learning
- action sets
- planning under uncertainty
- decision theoretic planning
- infinite horizon
- reward function
- reachability analysis
- partially observable
- average cost
- factored mdps
- stochastic games
- average reward
- learning algorithm
- markov decision process
- model free
- markov model
- state abstraction
- multi agent
- machine learning
- transition matrices
- sample size
- partially observable markov decision processes
- admissible heuristics
- control problems
- heuristic search
- search algorithm