Hindsight Foresight Relabeling for Meta-Reinforcement Learning.
Michael WanJian PengTanmay GangwaniPublished in: CoRR (2021)
Keyphrases
- reinforcement learning
- action selection
- reinforcement learning algorithms
- state space
- meta level
- multi agent
- machine learning
- function approximation
- markov decision processes
- learning algorithm
- optimal policy
- model free
- control problems
- markov decision process
- high level abstraction
- robotic control
- autonomous learning
- temporal difference learning
- supervised learning
- temporal difference
- learning process
- dynamic programming
- artificial neural networks
- objective function
- robot control
- search space
- stochastic approximation
- multi agent reinforcement learning
- relational reinforcement learning
- meta reasoning
- graph transformation
- domain knowledge