Leveraging unlabeled data to predict out-of-distribution performance.
Saurabh GargSivaraman BalakrishnanZachary Chase LiptonBehnam NeyshaburHanie SedghiPublished in: ICLR (2022)
Keyphrases
- unlabeled data
- semi supervised learning
- labeled data
- semi supervised
- active learning
- co training
- semi supervised classification
- supervised learning
- training data
- training and test data
- data points
- labeled examples
- labeled and unlabeled data
- text classification
- training set
- text categorization
- labeled training data
- learning algorithm
- class labels
- number of labeled examples
- transfer learning
- domain adaptation
- training examples
- prior knowledge
- small set of labeled
- positive examples
- sample selection bias
- label propagation
- data sets
- supervised and semi supervised
- decision boundary
- instance selection
- supervised learning algorithms