Version Space Support Vector Machines
Evgueni Nikolaevich Smirnov, Ida G. Sprinkhuizen-Kuyper, Georgi I. Nalbantov, Stijn Vanderlooy · 2006
Abstract. We argue to use version spaces as an approach to reliable classification. The key idea is to construct version spaces containing the hypotheses of the target concept or of its close approximations. As a result the unanimous-voting classification rule of version spaces does not misclassify; i.e., instance classifications become reliable. We propose to implement version spaces using support vector machines. The resulting combination is called version space support vector machines (VSSVMs). Experiments show that VSSVMs are able to outperform the existing approaches to reliable classification. 1