Convex Surrogates for Unbiased Loss Functions in Extreme Classification With Missing Labels.
Mohammadreza QaraeiErik SchultheisPriyanshu GuptaRohit BabbarPublished in: WWW (2021)
Keyphrases
- loss function
- hinge loss
- partially labeled data
- risk minimization
- pairwise
- support vector
- empirical risk
- pairwise constraints
- class labels
- convex optimization
- reproducing kernel hilbert space
- logistic regression
- loss minimization
- semi supervised
- training set
- image classification
- decision boundary
- binary classification
- convex loss functions
- support vector machine svm
- support vector machine
- feature space
- precision and recall
- potential functions
- text classification
- squared error
- multi class
- bayes rule
- learning to rank
- svm classifier
- class conditional
- supervised learning
- soft margin
- feature vectors
- training data