Nonlinear Regularization Path for Quadratic Loss Support Vector Machines.
Masayuki KarasuyamaIchiro TakeuchiPublished in: IEEE Trans. Neural Networks (2011)
Keyphrases
- support vector
- learning machines
- large margin classifiers
- sequential quadratic programming
- solution path
- quadratic program
- loss function
- pairwise
- taylor series expansion
- kernel methods
- kernel function
- penalty functions
- square loss
- polynomial kernels
- computational complexity
- logistic regression
- svm classifier
- maximum margin
- binary classification
- regularization parameter
- kernel machines
- penalized likelihood
- prior information
- data dependent
- feature selection
- objective function
- multi class
- cross validation
- generalization ability
- mercer kernels
- high dimensional feature space
- reproducing kernel hilbert space
- learning problems
- support vectors
- piecewise linear
- shortest path
- radial basis function
- support vector machine
- data points
- classification accuracy
- high dimensional
- training data
- neural network
- hyperparameters
- learning algorithm