On the Design of Loss Functions for Classification: theory, robustness to outliers, and SavageBoost.
Hamed Masnadi-ShiraziNuno VasconcelosPublished in: NIPS (2008)
Keyphrases
- loss function
- hinge loss
- support vector
- squared error
- loss minimization
- pairwise
- classification accuracy
- support vector machine
- logistic regression
- convex loss functions
- boosting algorithms
- benchmark datasets
- feature extraction
- decision trees
- density estimation
- generalization ability
- machine learning algorithms
- bregman divergences
- multi class
- update rules
- feature selection