On Generalization Performance and Non-Convex Optimization of Extended υ-Support Vector Machine.

The υ-support vector classification (υ-SVC) algorithm was shown to work well and provide intuitive interpretations, e.g., the parameter v roughly specifies the fraction of support vectors. Although ii corresponds to a fraction, it cannot take the entire range between 0 and 1 in its original form. This problem was settled by a non-convex extension of υ-SVC and the extended method was experimentally shown to generalize better than original υ-SVC. However, its good generalization performance and convergence properties of the optimization algorithm have not been studied yet. In this paper, we provide new theoretical insights into these issues and propose a novel υ-SVC algorithm that has guaranteed generalization performance and convergence properties. [ABSTRACT FROM AUTHOR]