Off-policy Maximum Entropy Reinforcement Learning : Soft Actor-Critic with Advantage Weighted Mixture Policy(SAC-AWMP).
Zhimin HouKuangen ZhangYi WanDongyu LiChenglong FuHaoyong YuPublished in: CoRR (2020)
Keyphrases
- maximum entropy
- actor critic
- reinforcement learning
- policy gradient
- temporal difference
- approximate dynamic programming
- optimal control
- reinforcement learning algorithms
- gradient method
- policy iteration
- function approximation
- markov models
- neuro fuzzy
- policy gradient methods
- conditional random fields
- global optimization
- model free
- optimal policy
- average reward
- rl algorithms
- state space
- markov decision processes
- partially observable
- step size
- linear program
- multi agent
- action selection
- single agent
- transfer learning
- natural actor critic
- function approximators
- supervised learning
- markov decision process
- dynamic programming
- semi supervised
- stochastic games
- action space
- linear programming
- learning problems