Asymmetric semi-supervised boosting for SVM active learning in CBIR.
Jun WuZheng-Kui LinMingyu LuPublished in: CIVR (2010)
Keyphrases
- semi supervised
- active learning
- partially labeled data
- relevance feedback
- kernel machines
- semi supervised support vector machines
- cost sensitive
- svm classifier
- fully labeled
- batch mode
- boosting framework
- learning algorithm
- labeled data
- semi supervised learning
- training set
- support vector machine svm
- support vector
- feature selection
- machine learning
- image retrieval
- unlabeled data
- support vector machine
- unlabeled instances
- supervised learning
- sample selection
- adaboost algorithm
- training process
- ensemble classifier
- labeled examples
- binary classification
- image database
- pool based active learning
- pairwise
- base classifiers
- active learning framework
- multi class
- pairwise constraints
- ensemble learning
- feature vectors
- training examples
- learning process
- visual features
- decision stumps
- kernel function
- medical image retrieval
- image content
- generalization ability
- multi class classification
- co training
- generalization error
- class imbalance
- training data
- feature space
- selective sampling
- cbir systems
- multiple instance learning
- face detection
- text classification
- instance level constraints
- random sampling
- support vectors
- kernel methods
- multiclass classification
- labeled and unlabeled data
- ensemble methods
- hyperplane