Support vector machines and strictly positive definite kernel: The regularization hyperparameter is more important than the kernel hyperparameters.
Luca OnetoAlessandro GhioSandro RidellaDavide AnguitaPublished in: IJCNN (2015)
Keyphrases
- hyperparameters
- support vector
- positive definite
- kernel function
- kernel machines
- rbf kernel
- reproducing kernel hilbert space
- gaussian process
- cross validation
- model selection
- kernel methods
- kernel matrix
- gaussian processes
- learning machines
- loss function
- support vector machine
- bayesian inference
- random sampling
- svm classifier
- ls svm
- prior information
- em algorithm
- support vector regression
- covariance matrix
- support vectors
- diffusion tensor
- graph kernels
- input space
- geometric structure
- kernel learning
- feature selection
- incomplete data
- noise level
- maximum a posteriori
- closed form
- high dimensional
- semi parametric
- feature space
- incremental learning
- bayesian framework
- sample size
- multiple kernel learning
- support vector machine svm
- maximum likelihood
- active learning
- lower bound
- machine learning