Learning Action Translator for Meta Reinforcement Learning on Sparse-Reward Tasks.
Yijie GuoQiucheng WuHonglak LeePublished in: AAAI (2022)
Keyphrases
- reinforcement learning
- learning algorithm
- learning process
- multiple tasks
- supervised learning
- action selection
- eligibility traces
- learning agent
- learning problems
- multi task learning
- partially observable environments
- reward shaping
- learning agents
- state action
- learning capabilities
- function approximation
- learning tasks
- transfer learning
- online learning
- state space
- multi agent
- dictionary learning
- multi task
- reinforcement learning algorithms
- complex domains
- optimal policy
- policy gradient
- inverse reinforcement learning
- dynamic programming
- partially observable domains