Joint semi-supervised and contrastive learning enables zero-shot domain-adaptation and multi-domain segmentation.
Alvaro GomarizYusuke KikuchiYun Yvonna LiThomas AlbrechtAndreas MaunzDaniela FerraraHuanxiang LuOrcun GokselPublished in: CoRR (2024)
Keyphrases
- domain adaptation
- semi supervised
- cross domain
- multi domain
- labeled data
- semi supervised learning
- supervised learning
- unlabeled data
- covariate shift
- co training
- unsupervised learning
- labeled and unlabeled data
- active learning
- prior knowledge
- pairwise
- machine learning
- general purpose
- learning tasks
- multiple sources
- reinforcement learning
- multi view
- co occurrence
- text categorization
- natural language processing
- learning algorithm