Having covered the foundations, in this part of the book we dive into machine learning in a hands-on fashion, armed with TensorFlow.js. We start in chapter 2 with a simple machine-learning task—regression (predicting a single number)—and work toward more sophisticated tasks such as binary and multiclass classification in chapters 3 and 4. In lockstep with task types, you’ll also see a gentle progression from simple data (flat arrays of numbers) to more complex ones (images and sounds). The mathematical underpinning of methods such as backpropagation will be introduced alongside concrete problems and the code that solves them. We eschew formal math in favor of more intuitive explanations, diagrams, and pseudo-code. Chapter 5 discusses transfer learning, an efficient reuse of pretrained neural networks to adapt to new data, and presents an approach especially suited to the deep-learning browser environment.