Dyna-MLAC: Trading Computational and Sample Complexities in Actor-Critic Reinforcement Learning.
Bruno CostaWouter CaarlsDaniel Sadoc MenaschéPublished in: BRACIS (2015)
Keyphrases
- actor critic
- reinforcement learning
- temporal difference
- function approximation
- rl algorithms
- policy gradient
- reinforcement learning algorithms
- optimal control
- approximate dynamic programming
- temporal difference learning
- neuro fuzzy
- reinforcement learning methods
- function approximators
- gradient method
- model free
- policy iteration
- state space
- evaluation function
- policy gradient methods
- learning problems
- monte carlo
- markov decision processes
- action selection
- step size
- sample size
- multi agent
- control problems
- average reward
- dynamic programming
- search space
- adaptive control
- machine learning