Approximate Newton Methods for Policy Search in Markov Decision Processes.
Thomas FurmstonGuy LeverDavid BarberPublished in: J. Mach. Learn. Res. (2016)
Keyphrases
- markov decision processes
- policy search
- reinforcement learning algorithms
- reinforcement learning
- reward function
- continuous state
- partially observable markov decision processes
- action space
- dynamic programming
- finite state
- state space
- markov decision problems
- optimal policy
- policy iteration
- partially observable
- average cost
- stochastic games
- planning under uncertainty
- policy gradient
- markov games
- average reward
- decision processes
- markov decision process
- learning algorithm
- finite horizon
- infinite horizon
- risk sensitive
- search space