Does Learning from Decentralized Non-IID Unlabeled Data Benefit from Self Supervision?
Lirui WangKaiqing ZhangYunzhu LiYonglong TianRuss TedrakePublished in: ICLR (2023)
Keyphrases
- unlabeled data
- active learning
- learning algorithm
- supervised learning
- semi supervised learning
- labeled data
- labeled and unlabeled data
- positive examples
- semi supervised
- co training
- domain adaptation
- sample selection bias
- supervised learning algorithms
- learning process
- semi supervised classification
- training data
- prior knowledge
- partially labeled data
- number of labeled examples
- class labels
- labeled instances
- learning problems
- learning tasks
- transfer learning
- data sets
- unsupervised learning
- text classification
- training set
- negative examples
- text categorization
- maximum likelihood
- multi view learning
- learning frameworks
- feature space
- data analysis
- small set of labeled