Y. Yun (Japan), M. Yoon (Korea), T. Asada, and H. Nakayama (Japan)
support vector classification, surplus variable, soft margin, total margin, -SVC
In pattern classification problems with two-class sets, Sup port Vector Machines (SVMs) have shown good perfor mance in finding a linear classification function which dis criminate well unknown example points. This paper intro duces surplus variables for correctly classified data points as a measure of the distance between each data point and the separating hyperplane. A new algorithm for finding op timal separating hyperplanes is proposed by maximizing the surplus variables and minimizing the slack variables. The proposed method is compared with several algorithms through numerical examples.
Important Links:
Go Back
