Fine-tuning or top-tuning? Transfer learning with pretrained features and fast kernel methods.
Paolo Didier AlfanoVito Paolo PastoreLorenzo RosascoFrancesca OdonePublished in: CoRR (2022)
Keyphrases
- fine tuning
- kernel methods
- transfer learning
- learning tasks
- feature space
- machine learning
- kernel function
- learning problems
- support vector
- active learning
- support vector machine
- labeled data
- feature set
- knowledge transfer
- cross domain
- fine tuned
- learning algorithm
- feature vectors
- kernel matrix
- multi task
- transfer knowledge
- machine learning algorithms
- k nearest neighbor
- text classification
- svm classifier
- semi supervised learning
- domain adaptation
- multi task learning
- multiple kernel learning
- kernel learning
- collaborative filtering
- classification accuracy
- feature extraction