Stochastic smoothing of the top-K calibrated hinge loss for deep imbalanced classification.
Camille GarcinMaximilien ServajeanAlexis JolyJoseph SalmonPublished in: ICML (2022)
Keyphrases
- hinge loss
- loss function
- decision boundary
- potential functions
- convex optimization
- risk minimization
- multi class
- soft margin
- pattern classification
- feature extraction
- reproducing kernel hilbert space
- support vector machine svm
- class imbalance
- machine learning
- support vector machine
- classification accuracy
- decision trees
- binary classification
- class labels
- kernel function
- supervised learning
- training set
- density estimation
- machine learning algorithms
- pairwise
- training data