A decentralized policy gradient approach to multi-task reinforcement learning.
Sihan ZengMalik Aqeel AnwarThinh T. DoanArijit RaychowdhuryJustin RombergPublished in: UAI (2021)
Keyphrases
- multi task
- policy gradient
- reinforcement learning
- actor critic
- function approximation
- transfer learning
- learning tasks
- learning problems
- reinforcement learning algorithms
- multi agent
- optimal control
- gradient method
- metric learning
- multi class
- reinforcement learning methods
- feature selection
- function approximators
- state space
- machine learning
- learning algorithm
- temporal difference
- model free
- single agent
- partially observable markov decision processes
- approximation methods
- rl algorithms
- average reward
- supervised learning
- state action
- variance reduction
- learning process
- learning experience
- partially observable
- optimal policy
- active learning
- neural network