Exploration Bonus for Regret Minimization in Undiscounted Discrete and Continuous Markov Decision Processes.
Jian QianRonan FruitMatteo PirottaAlessandro LazaricPublished in: CoRR (2018)
Keyphrases
- markov decision processes
- continuous state spaces
- regret minimization
- model based reinforcement learning
- action space
- state space
- policy iteration
- interval estimation
- finite state
- optimal policy
- continuous state
- average reward
- reinforcement learning
- dynamic programming
- reachability analysis
- reinforcement learning algorithms
- decision theoretic planning
- stochastic games
- average cost
- planning under uncertainty
- infinite horizon
- transition matrices
- markov decision process
- markov decision problems
- partially observable
- nash equilibrium
- decision processes
- game theoretic
- machine learning
- finite number
- least squares
- search space
- decision making