Optimizing the Classification Cost using SVMs with a Double Hinge Loss.
Ahmed AmirouDjaffar Ould AbdeslamZahia ZidelmalA. MohamedJean MerckléPublished in: Informatica (Slovenia) (2014)
Keyphrases
- hinge loss
- loss function
- support vector
- soft margin
- decision boundary
- risk minimization
- potential functions
- cost sensitive
- convex optimization
- support vector machine svm
- support vector machine
- multi class
- binary classification
- reproducing kernel hilbert space
- feature selection
- svm classifier
- machine learning
- pairwise constraints
- line search
- generalization ability
- pairwise
- training set
- pattern classification
- decision trees
- support vectors
- image classification
- generalization error
- supervised learning
- exact inference
- naive bayes classifier
- kernel function
- class labels
- feature extraction
- linearly separable
- kernel methods
- training data
- boosting algorithms
- maximum margin
- machine learning algorithms
- classification accuracy
- hyperplane