Hindsight Foresight Relabeling for Meta-Reinforcement Learning.
Michael WanJian PengTanmay GangwaniPublished in: ICLR (2022)
Keyphrases
- reinforcement learning
- action selection
- function approximation
- reinforcement learning algorithms
- state space
- model free
- meta level
- learning algorithm
- dynamic programming
- high level abstraction
- meta reasoning
- optimal policy
- control policy
- temporal difference
- search engine
- multi agent reinforcement learning
- direct policy search
- learning agents
- real time
- action space
- control problems
- transfer learning
- markov chain
- supervised learning
- information retrieval