Exploiting Class Similarity for Machine Learning with Confidence Labels and Projective Loss Functions.
Gautam Rajendrakumar GareJohn Michael GaleottiPublished in: CoRR (2021)
Keyphrases
- loss function
- machine learning
- pairwise
- reproducing kernel hilbert space
- learning models
- support vector
- class labels
- similarity measure
- squared error
- loss minimization
- learning to rank
- bregman divergences
- machine learning algorithms
- partially labeled data
- learning to rank algorithms
- feature selection
- training data
- learning tasks
- training set
- distance measure
- pairwise constraints
- multi class classification
- distance function
- active learning
- potential functions
- supervised learning
- model selection
- special case
- precision and recall
- risk minimization
- hinge loss
- update rules
- decision trees
- training examples