The Reactor: A fast and sample-efficient Actor-Critic agent for Reinforcement Learning.
Audrunas GruslysWill DabneyMohammad Gheshlaghi AzarBilal PiotMarc G. BellemareRémi MunosPublished in: ICLR (Poster) (2018)
Keyphrases
- reinforcement learning
- actor critic
- function approximation
- approximate dynamic programming
- multi agent
- temporal difference
- multi agent systems
- policy gradient
- reinforcement learning algorithms
- action selection
- state space
- learning algorithm
- optimal control
- neural network
- model free
- neuro fuzzy
- linear program
- control problems
- state action
- markov decision processes
- dynamic environments
- policy gradient methods