Although the basic formulation of Support Vector Machines is for binary
classifiers, i.e., those with only two classes, they can be used for classification into
multiple classes as follows: If there are N classes, we build N classifiers, with
classifier i performing a binary classification, classifying a point either as in class
i or not in class i. Given a point, each classifier i also outputs a value indicating
how related a given point is to class i. We then apply all N classifiers on a given
point, and choose the class for which the relatedness value is the highest.