Warm Start Active Learning with Proxy Labels and Selection via Semi-supervised Fine-Tuning.
Vishwesh NathDong YangHolger R. RothDaguang XuPublished in: MICCAI (8) (2022)
Keyphrases
- fine tuning
- semi supervised
- active learning
- label propagation
- unlabeled examples
- labeled examples
- semi supervised learning
- labeled data
- partially labeled data
- annotation effort
- unlabeled samples
- pairwise constraints
- training examples
- unlabeled data
- fully supervised
- unlabeled instances
- label information
- viable alternative
- supervised learning
- pairwise
- partially labeled
- fine tune
- training set
- unsupervised learning
- learning algorithm
- label noise
- pool based active learning
- fine tuned
- co training
- labeling effort
- training data
- labeled samples
- fully labeled
- machine learning
- active learning framework
- multi view
- learning strategies
- class labels
- weakly supervised
- transfer learning
- batch mode
- incremental learning
- labeled and unlabeled data
- learning process
- relevance feedback
- semi supervised classification
- data sets
- selective sampling
- general purpose
- labeled instances
- random sampling
- semi supervised clustering
- multiple instance learning
- selection strategy
- decision boundary