Off-Dynamics Reinforcement Learning: Training for Transfer with Domain Classifiers.
Benjamin EysenbachShreyas ChaudhariSwapnil AsawaSergey LevineRuslan SalakhutdinovPublished in: ICLR (2021)
Keyphrases
- reinforcement learning
- transfer learning
- training examples
- test set
- training set
- training samples
- cross domain
- labeled training data
- supervised learning
- domain specific
- function approximation
- labeled data
- classifier training
- support vector
- multi layer perceptron
- feature selection
- training data
- decision trees
- partially observable domains
- linear svm
- discriminative classifiers
- learning classifier systems
- domain independent
- multi class
- knowledge transfer
- optimal control
- target domain
- training dataset
- linear classifiers
- unlabeled data
- temporal difference
- semi supervised learning
- semi supervised
- state space
- supervised training
- classification performances
- deterministic domains
- active learning