Learning Deep Learning: Theory and Practice of Neural Networks, Computer Vision, NLP, and Transformers using TensorFlow

Chapter 2

Gradient-Based Learning

In this chapter, we describe how the perceptron learning algorithm works, which we then build upon in Chapter 3, “Sigmoid Neurons and Backpropagation,” by extending it to multilevel networks. These two chapters contain more mathematical content than other chapters in this book, but we also describe the concepts in an intuitive manner for readers who do not like reading mathematical formulas.

Intuitive Explanation of the Perceptron Learning Algorithm

In Chapter 1, “The Rosenblatt Perceptron,” we presented and used the perceptron learning algorithm, but we did not explain why it works. Let us now look at what the learning algorithm does. To refresh our memory, the weight adjustment step in the perceptron learning ...