Active Learning of Very-High Resolution Optical Imagery with SVM: Entropy vs Margin Sampling.
Devis TuiaFrédéric RatleFabio PacificiAlexei PozdnoukhovMikhail F. KanevskiFabio Del FrateDomenico SoliminiWilliam J. EmeryPublished in: IGARSS (4) (2008)
Keyphrases
- active learning
- support vector
- random sampling
- training set
- sample selection
- sampling strategies
- maximum margin
- generalization error
- decision boundary
- electro optical
- support vector machine svm
- imbalanced data
- training examples
- support vector machine
- statistical learning theory
- support vectors
- standard svm
- active sampling
- active learning framework
- sampling algorithm
- stratified sampling
- svm classifier
- selective sampling
- class imbalance
- unlabeled data
- training data
- imbalanced class distribution
- soft margin
- high resolution
- cost sensitive
- semi supervised
- imaging sensors
- information theory
- svm classification
- machine learning
- information theoretic
- labeled data
- kernel function
- svm training
- minority class
- feature vectors
- supervised learning
- sampling methods
- sampling strategy
- learning algorithm
- binary classification
- kernel methods
- multi class
- knn
- satellite imagery
- large margin classifiers
- decision function
- mutual information
- classification algorithm
- feature selection
- image data
- learning process
- expected loss
- imbalanced datasets
- semi supervised learning
- infrared
- computer vision
- generalization ability
- hyperplane
- active learning strategies
- data sets
- decision trees
- pairwise
- cross validation
- linear classifiers
- information gain