Introduction to Support Vector Machines

Abstract

Support Vector Machines (SVMs) have been used as an efficient clas- sification tool in the last two decades. SVMs are based on statistical learning theory. The formulation of SVM embodies the Structural Risk Minimisation (SRM) principle. Before SVM, the conventional neural net- works used Empirical RiskMinimisation (ERM). But SRMproved to give superior results than its predecessors. SVM gives global optimum solu- tion by solving a convex quadratic programming problem. It has good generalization ability and can efficiently develop the classifier model in lesser time than neural networks or other probability based models. This chapter explains the formulation of a classical SVM, and introduces the terminology associated with SVM.