In this chapter, we discussed the support vector machine—a powerful model that can mitigate some of the limitations of perceptrons. The perceptron can be used effectively for linearly separable classification problems, but it cannot express more complex decision boundaries without expanding the feature space to higher dimensions. Unfortunately, this expansion is prone to computation and generalization problems. Support vector machines redress the first problem using kernels, which avoid explicitly computing the feature mapping. They redress the second problem by maximizing the margin between the decision boundary and the nearest instances. In the next chapter, we will discuss models called artificial neural networks, which, like support vector machines, extend the perceptron to overcome its limitations.