Decoupling Exploration and Exploitation for Meta-Reinforcement Learning without Sacrifices.
Evan Zheran LiuAditi RaghunathanPercy LiangChelsea FinnPublished in: ICML (2021)
Keyphrases
- exploration exploitation tradeoff
- reinforcement learning
- function approximation
- exploration strategy
- objective function
- active exploration
- relevance feedback
- model based reinforcement learning
- input output
- exploration exploitation
- action selection
- meta level
- markov decision processes
- reinforcement learning algorithms
- multi agent reinforcement learning
- multi agent
- autonomous learning
- model free
- learning algorithm
- dynamic programming
- stochastic approximation
- reinforcement learning methods
- learning classifier systems
- meta reasoning
- learning problems
- state space
- robotic control
- case study