ABSTRACT
While the origins of support vector machines are old, their modern treatment was initiated in Boser et al. and Cortes and Vapnik (binary classification) and Drucker et al. (regression). As is often the case in machine learning, it is easier to explain a complex tool through an illustration with binary classification. In fact, sometimes, it is originally how the tool was designed e.g., for the perceptron. The interesting features of the scheme are those that we have not mentioned yet, that is, the grey dotted lines. These lines represent the no-man’s land in which no observation falls when the green model is enforced. The two margins are computed as the parallel lines that maximize the distance between the model and the closest points that are correctly classified. These points are called support vectors, which justifies the name of the technique.
