Incremental Twin Support Vector Machines

• Published in: Modeling, Computation And Optimization pp. 263 - 272
• Main Authors: Khemchandani, Reshma, Jayadeva, Chandra, Suresh
• Format: Book Chapter
• Language: English
• Published: WORLD SCIENTIFIC 01.04.2009
• Subjects: Support vector machines; pattern classification; machine learning; incremental learning
• ISBN: 9789814467896; 9814273511; 9789814273503; 9814467898; 9789814273510; 9814273503
• DOI: 10.1142/9789814273510_0017

Support Vector Machines (SVMs) suffer from the problem of large memory requirements and CPU time when trained in batch mode on large data sets. Therefore incremental techniques have been developed to facilitate batch SVM learning. In this chapter we propose a new incremental technique called Incremental Twin Support Vector Machines for training in batch mode. This technique is based on a newly developed classifier, called Twin Support Vector Machines (TWSVM) classifier. The TWSVM classifier determines two non-parallel planes by solving two related support vector machines-type problems, each of which is smaller than in a conventional Incremental SVM. Numerical implementation on several benchmark datasets has shown that the Incremental Twin SVM is not only fast, but also has good generalization.