Nonlinear regularization path for the modified Huber loss Support Vector Machines.
Masayuki KarasuyamaIchiro TakeuchiPublished in: IJCNN (2010)
Keyphrases
- support vector
- loss function
- solution path
- learning machines
- large margin classifiers
- robust statistics
- kernel machines
- quadratic program
- reproducing kernel hilbert space
- cross validation
- support vector machine
- maximum margin
- kernel function
- risk minimization
- logistic regression
- regularization parameter
- generalization ability
- shape from shading
- kernel methods
- classification accuracy
- binary classification
- decision function
- pairwise
- shortest path
- svm classifier
- stochastic gradient descent
- polynomial kernels
- neural network
- mercer kernels
- robust estimation
- hyperparameters
- hyperplane
- prior information
- radial basis function
- maximum a posteriori
- training data