Deterministic Policy Optimization by Combining Pathwise and Score Function Estimators for Discrete Action Spaces.
Daniel LevyStefano ErmonPublished in: CoRR (2017)
Keyphrases
- action space
- score function
- continuous state spaces
- continuous action
- continuous state
- state space
- policy search
- real valued
- markov decision processes
- state and action spaces
- interval valued
- reinforcement learning
- control policies
- state action
- optimal policy
- partially observable markov decision processes
- markov decision process
- action selection
- stochastic processes
- markov decision problems
- state dependent
- finite state
- fuzzy numbers
- supervised learning
- dynamic programming
- control system
- clustering algorithm