Directed Policy Gradient for Safe Reinforcement Learning with Human Advice.
Hélène PlisnierDenis SteckelmacherTim BrysDiederik M. RoijersAnn NowéPublished in: CoRR (2018)
Keyphrases
- policy gradient
- reinforcement learning
- actor critic
- function approximation
- policy search
- reinforcement learning algorithms
- optimal control
- policy gradient methods
- model free reinforcement learning
- gradient method
- variance reduction
- approximation methods
- function approximators
- model free
- markov decision processes
- reinforcement learning methods
- partially observable markov decision processes
- average reward
- temporal difference
- real valued
- supervised learning
- state space
- multi agent
- single agent
- learning problems
- approximate dynamic programming
- learning algorithm