Unified dual for bi-class SVM approaches

Abstract:

SVM theory was originally developed on the basis of a separable binary classification problem, and other approaches have been later introduced. In this paper, we demonstrated that all these approaches admit the same dual problem formulation. © 2005 Pattern Recognition Society. Published by Elsevier Ltd. All rights reserved.

Año de publicación:

2005

Keywords:

• Kernels
• SVM
• QP-problem
• Binary classification
• Large margin principle

Fuente:

scopus