COCA: Classifier-Oriented Calibration for Source-Free Universal Domain Adaptation via Textual Prototype.
Xinghong LiuYi ZhouTao ZhouChun-Mei FengLing ShaoPublished in: CoRR (2023)
Keyphrases
- domain adaptation
- target domain
- multiple sources
- training data
- source domain
- training samples
- test data
- cross domain
- co training
- sufficient training data
- transfer learning
- labeled data
- training and test data
- learning algorithm
- training set
- domain specific
- semi supervised
- support vector machine
- document classification
- data sets
- semi supervised learning
- unlabeled data
- feature selection
- training examples
- classification algorithm
- decision trees
- source and target domains
- covariate shift
- natural language
- machine learning
- databases
- single view
- sentiment classification
- prior knowledge
- support vector
- keywords