We'll begin this chapter by returning to a situation that should be very familiar by now: the binary classification task. Once again, we'll be thinking about the problem of how to design a model that will correctly predict whether an observation belongs to one of two possible classes. We've already seen that this task is simplest when the two classes are linearly separable, that is, when we can find a separating hyperplane (a plane in a multidimensional space) in the space of our features so that all the observations on one side of the hyperplane belong to one class and all the observations that lie on the other side belong to the second class. Depending on the structure, assumptions, and optimizing criterion that our particular model uses, we could end up with one of infinitely many such hyperplanes.
Let's visualize this scenario using some data in a two-dimensional feature space, where the separating hyperplane is just a separating line:
