Python Neural Network Programming

This book is the Chinese translation of Make Your Own Neural Network by the British technical author Tariq Rashid, translated by Lin Ci and published by Posts & Telecom Press in 2018 as part of the "Asynchronous Books · Deep Learning Series." Taking "write a neural network from scratch" as its throughline, the book deliberately avoids difficult mathematical derivations and explains, layer by layer and in language a secondary-school student can follow, how a neural network works. Chapter 1 begins with "a simple predicting machine," first building a linear classifier, then showing why a single classifier cannot solve complex problems, and from there introduces neurons, tracking signals between layers, what matrix multiplication is for, and how a multilayer network updates weights by back-propagating errors. Chapter 2 moves into practice, introducing the basic use of Python and IPython and leading readers step by step to build their own three-layer neural network in Python 3, then train and test it on the MNIST handwritten-digit dataset to check its recognition performance. Chapter 3 opens the "black box" of the trained network a little further, observing its weights and intermediate layers, trying it on one's own handwritten digits, and expanding the training data by rotating images. The appendixes add a primer on calculus and instructions for running the code on a Raspberry Pi. The book runs to about 200 pages and includes color illustrations; its emphasis is on getting hands-on code running rather than listing formulas, making it an introductory neural-network book suitable for complete beginners.