ProxyMix: Proxy-based Mixup Training with Label Refinery for Source-Free Domain Adaptation.
Yuhe DingLijun ShengJian LiangAihua ZhengRan HePublished in: CoRR (2022)
Keyphrases
- domain adaptation
- target domain
- multiple sources
- covariate shift
- source domain
- training samples
- cross domain
- transfer learning
- labeled data
- semi supervised
- supervised learning
- training process
- training data
- training set
- unlabeled data
- manifold alignment
- semi supervised learning
- training examples
- training and test data
- test data
- document classification
- class labels
- label information
- data sets
- general purpose
- co training
- sentiment classification
- multi label
- test set