What is SVM

The support vector machine, or SVM, is a parametric supervised learning model that can be used for both classification and regression. The SVM can be seen as an artificial neuron, specifically perceptron with sign activation, using a modified loss function to find an ideal decision boundary that maximizes the gap between classes. SVMs are one of the most efficient methods available, and are competitive with the current cutting edge. SVMs are exclusively binary classifiers, meaning that if one wishes to apply SVMs to multiclass classification problems, they must necessarily use an ensemble of SVMs.

This article introduces the history, fundamental theory, and usage of SVM methods. It focuses first on the classic hard-margin formulation of SVMs, and then moves into the more popular soft-margin. It then defines and explains the dual form of SVM, motivating the famous kernel SVM method.