Deep Learning with Python

This book is a beginner's guide to deep learning practice using Python. Its original title is Deep Learning Step by Step with Python: A Very Gentle Introduction to Deep Neural Networks for Practical Data Science. Rather than starting from a mass of formulas and derivations, it takes practical cases as its main thread, explaining in plain and accessible terms the two most common tasks of deep neural networks—classification and regression—and uses them to clarify several core issues in deep learning models, helping readers build a clear overall picture of deep learning. The book has 9 chapters. Chapter 1 explains how to read the book and prepare the environment (obtaining Python and common packages). Chapter 2 begins with "why learn deep learning," distinguishing supervised from unsupervised learning and explaining what deep learning can solve and which fields use it. Chapter 3 introduces neural network fundamentals, including a brief history, network topology, artificial neurons, activation functions (with sigmoid as an example), and gradient descent and stochastic gradient descent. Chapter 4 gives an overview of deep neural networks and methods for rapidly approximating arbitrary functions. Chapter 5 explains how to build customizable deep prediction models, covering prediction targets, data standardization, and splitting training and test sets. Chapter 6 discusses techniques for improving performance. Chapter 7 analyzes binary classification neural networks. Chapter 8 covers the way to build good models. Chapter 9 presents a simple application of deep neural networks to multiclass classification. The book runs about 154 pages and 133,000 Chinese characters, with a brisk pace suited to completing a hands-on introduction to deep learning in a short time.