The Perils of Learning From Unlabeled Data: Backdoor Attacks on Semi-supervised Learning.
Virat ShejwalkarLingjuan LyuAmir HoumansadrPublished in: ICCV (2023)
Keyphrases
- semi supervised learning
- unlabeled data
- labeled and unlabeled data
- labeled data
- supervised learning
- co training
- semi supervised
- learning problems
- active learning
- unsupervised learning
- learning models
- semi supervised classification
- multi view learning
- domain adaptation
- learning algorithm
- supervised learning algorithms
- machine learning
- training data
- learning frameworks
- labeled examples
- labeled instances
- semi supervised learning setting
- data points
- improve the classification accuracy
- class labels
- positive examples
- prior knowledge
- label propagation
- labeled training data
- partially labeled
- transfer learning
- partially labeled data
- semi supervised learning methods
- training examples
- learning process
- semi supervised clustering
- manifold regularization
- training set
- learning tasks
- multi view
- reinforcement learning
- transductive svm
- feature extraction
- unlabeled instances
- instance selection
- feature space
- semisupervised learning
- number of labeled examples
- abundant unlabeled data