SVM Classifiers – Concepts and Applications to Character Recognition
Alexander Thome · InTech eBooks · 2012
By nature SVMs are essentially binary classifiers, however, based on several researchers' contributions they were adapted to handle multiple classes cases.The two most common approaches used are the One-Against-All and One-Against-One techniques, but this scenario is still an ongoing research topic.In this chapter we briefly discuss some basic concepts on SVM, describe novel approaches proposed in the literature and discuss some experimental tests applied to character recognition.The chapter is divided into 4 sections.Section 2 presents the theoretical aspects of the Support Vector Machines.Section 3 reviews some strategies to deal with multiple classes.Section 4 details some experiments on the usage of One-Against-All and One-Against-One approach applied to character recognition. Theoretical foundations of the SVMSupport vector machines are computational algorithms that construct a hyperplane or a set of hyperplanes in a high or infinite dimensional space.SVMs can be used for classification, regression, or other tasks.Intuitively, a separation between two linearly separable classes is achieved by any hyperplane that provides no misclassification on all data points of any of the considered classes, that is, all points belonging to class A are labeled as +1, for example, and all points belonging to class B are labeled as -1.