Robustness of learning algorithms using hinge loss with outlier indicators.
Takafumi KanamoriShuhei FujiwaraAkiko TakedaPublished in: Neural Networks (2017)
Keyphrases
- learning algorithm
- hinge loss
- loss function
- risk minimization
- binary classification
- active learning
- learning problems
- machine learning
- soft margin
- multi class
- machine learning algorithms
- boosting algorithms
- outlier detection
- learning tasks
- generalization ability
- decision boundary
- potential functions
- supervised learning
- pairwise
- bayesian networks
- convex optimization
- transfer learning
- maximum margin
- back propagation
- training examples