Deformation of log-likelihood loss function for multiclass boosting.
Takafumi KanamoriPublished in: Neural Networks (2010)
Keyphrases
- log likelihood
- loss function
- boosting algorithms
- multi class
- pairwise
- base classifiers
- multiclass classification
- cost sensitive
- maximum likelihood
- density estimation
- support vector
- information theoretic
- hinge loss
- multi class boosting
- em algorithm
- support vector machine
- soft margin
- base learners
- feature selection
- risk minimization
- closed form
- binary classification
- multi class classification
- semi supervised
- perceptron algorithm
- ensemble methods
- binary classifiers
- machine learning methods
- mutual information
- bayesian networks