In the previous chapter we achieved our goal of building a computer vision system—but only a basic one. The program we built is a binary classifier, as it assigns data to one of two classes: “5” and “not 5.” Now we’re going to push that program further and tackle the final challenge of Part I of this book: recognize any digit in MNIST.
Instead of binary classification, the problem of recognizing digits involves many classes—and for that reason, it’s called a multiclass classification problem. Don’t fret about multiclass classification, because it’s well within our grasp. In fact, here’s a simple recipe for multiclass classification: build a binary classifier for each class, and then combine the classifiers from each class into one multiclass classifier. Let’s turn that idea into code. Hold on tight: we’re tantalizingly close to our goal of building a full-fledged MNIST classifier.