Marginal Policy Gradients: A Unified Family of Estimators for Bounded Action Spaces with Applications.
Carson EisenachHaichuan YangJi LiuHan LiuPublished in: ICLR (Poster) (2019)
Keyphrases
- action space
- state space
- markov decision processes
- state and action spaces
- reinforcement learning
- real valued
- control policies
- action selection
- continuous state
- continuous state spaces
- stochastic processes
- reinforcement learning problems
- optimal policy
- markov decision process
- probability distribution
- markov decision problems
- state action
- function approximators
- skill learning
- single agent
- asymptotically optimal
- finite state
- heuristic search
- steady state
- special case
- reinforcement learning algorithms
- dynamic programming
- bayesian networks