Low-Precision Reinforcement Learning: Running Soft Actor-Critic in Half Precision.
Johan BjörckXiangyu ChenChristopher De SaCarla P. GomesKilian Q. WeinbergerPublished in: ICML (2021)
Keyphrases
- actor critic
- reinforcement learning
- temporal difference
- policy gradient
- reinforcement learning algorithms
- approximate dynamic programming
- optimal control
- function approximation
- gradient method
- neuro fuzzy
- state space
- policy iteration
- control problems
- policy gradient methods
- temporal difference learning
- model free
- monte carlo
- multi agent
- learning problems
- rl algorithms
- dynamic programming
- learning algorithm
- step size
- action selection
- evaluation function
- linear program
- partially observable markov decision processes
- markov decision processes
- optimal policy
- natural actor critic
- machine learning