Behavior-Guided Actor-Critic: Improving Exploration via Learning Policy Behavior Representation for Deep Reinforcement Learning.
Ammar FayadMajd IbrahimPublished in: CoRR (2021)
Keyphrases
- actor critic
- reinforcement learning
- policy gradient
- temporal difference
- approximate dynamic programming
- action selection
- optimal control
- policy gradient methods
- neuro fuzzy
- gradient method
- learning process
- exploration exploitation tradeoff
- optimal policy
- reinforcement learning algorithms
- learning algorithm
- machine learning
- function approximation
- markov decision processes
- supervised learning
- rl algorithms
- average reward
- policy iteration
- learning tasks
- function approximators
- autonomous robots
- temporal difference learning
- sufficient conditions
- learning problems
- state space
- reinforcement learning problems
- dynamic programming
- model free