Kernel machines such as the support vector machine are reviewed. Most of them are not suffered from the local optimum problem, because they are basically linear machines. Moreover, regularization technique provides them with high generalization performance as well as considerable accuracy.
展开▼
