Back in Chapter 2, Your First Classifier, we constructed a simple neural network for our character recognition effort. We ended the chapter with commendable mid-80% accuracy. Good start, but we can do much better!
In this chapter, we will retrofit our earlier classifier with far more powerful network architecture. Then, we'll delve into a much more difficult problem—handling color images from the CIFAR-10 dataset. The images will be much more difficult (cats, dogs, airplanes, and so on), so we'll bring more powerful tools to the table—specifically, a convolutional neural network. Let's begin.
