Pre-training as Batch Meta Reinforcement Learning with tiMe.
Quan VuongShuang LiuMinghua LiuKamil CiosekHao SuHenrik Iskov ChristensenPublished in: CoRR (2019)
Keyphrases
- reinforcement learning
- batch mode
- function approximation
- training set
- neural network
- training samples
- supervised learning
- model free
- real time
- information systems
- training phase
- test set
- markov decision processes
- training algorithm
- robotic control
- active learning
- temporal difference
- training process
- learning algorithm
- optimal policy
- markov chain
- knowledge base
- semi supervised
- state space