Generalizing Version Space Support Vector Machines for Non-Separable Data

Although version space support vector machines (VSSVMs) are a successful approach to reliable classification [6], they are restricted to separable data. This paper proposes generalized VSSVMs (GVSSVMs) applicable for separable and non-separable data. We show that GVSSVMs can outperform existing reliable-classification approaches.

• Type: Conference
• Year: 2006
• Where: ICDM
• Authors: Evgueni N. Smirnov, Ida G. Sprinkhuizen-Kuyper, Nikolay I. Nikolaev
• Keywords: Data Mining, ICDM 2006, Separable Data, Space Support Vector, Successful Approach