Together, we have witnessed the rapid development of machine learning and artificial intelligence in recent years, with important technological breakthroughs emerging one after another. Smartphones now not only understand what we say, but can even translate it into multiple languages. Self-driving cars are approaching the level of human driving in terms of safety; Computers can even diagnose some diseases more accurately and quickly than experienced doctors.

Weiqi originated in China and has a history of more than 3 000 years. Although the rules of Go are simpler than those of chess, the complexity of its long-term strategy far exceeds that of chess. In recent years, researchers have developed machine learning systems that have repeatedly beaten human go world champions. Not only that, but the system has even taught itself a new strategy that no one has figured out in 3,000 years! The fact that computers discover new strategies as they learn to complete a task is a major achievement in the entire field of machine learning

Young GAN

Compared with decades of research and accumulation of traditional neural networks, GAN came to prominence after a paper was published by Ian Goodfellow in 2014. This means that GAN research is just beginning, and there is unlimited room for creativity and exploration. It also means that we do not yet fully understand how to train gans as well as traditional neural networks. GAN can be very effective if done correctly. Most of the time, however, GAN doesn’t work. Today, many researchers are studying how GAN works and why it fails.

The neural network

GAN is made up of neural networks. While this book is a refresher, I recommend another of my Own, Make Your Own Neural Network.

Neural Network Programming in Python

It specializes in introducing neural networks and their operating principles and is very suitable for beginners. Meanwhile, it also includes calculus, gradient descent and other contents.

This book will take you on a fun yet methodical journey, starting with a very simple idea and gradually understanding how neural networks work. You don’t need to know any math beyond high school, and this book provides an easy-to-understand introduction to calculus. The goal of this book is to make neural networks accessible to as many general readers as possible. You will learn to develop your own neural network in Python and train it to recognize handwritten numbers, even as well as professional neural networks.

This book is for those who want to learn about deep learning, artificial intelligence, and neural networks, and especially for those who want to develop neural networks through Python programming.

PyTorch Generative Versus Network Programming is a good book to read if you want a preliminary understanding of gans and how they work, or if you are someone who builds gans using industrial-grade software.

PyTorch Generation vs. Network Programming

Generative Adversarial Network (GAN) is a new phenomenon in the field of neural networks and has been hailed as “the coolest idea in machine learning in the last 20 years.”

This book introduces the reader to generative adversarial networks in a straightforward, short way and teaches the reader how to write generative adversarial networks in a step-by-step manner using PyTorch. The book consists of three chapters and five appendices, which respectively introduce the basic knowledge of PyTorch, the development of neural network with PyTorch, the improvement of neural network to improve the effect, the introduction of CUDA and GPU to accelerate GAN training, and the generation of high quality image convolution GAN, conditional GAN and other topics. The appendix covers topics that have been neglected in many machine learning-related tutorials, including calculating ideal loss values for balanced GAN, probability distributions, and sampling, and how convolution works. It also briefly explains why gradient descent is not suitable for adversarial machine learning.

Once you’ve finished reading PyTorch Generation Versus Network Programming, you should be able to understand GAN and build a simple GAN with your own hands.

For more complex concepts, the book will try to use plain language with extensive illustrations to explain them. This book tries to avoid unnecessary terminology and mathematical formulas.

The goal of this book is to help readers from different backgrounds understand gans and build gans themselves.

This book is not an encyclopedia of GAN and cannot cover all aspects of GAN. We purposefully selected the best parts, enough to prepare the reader for further research.

For students who are taking machine learning-related courses, this book can help them get a quick start and lay the foundation for the rest of their study.

How to Use this book

The best way to learn a skill is to do it yourself. For this reason, concepts and theories are explained through a step-by-step, hands-on approach.

Even if the reader follows the instructions carefully, he or she may still encounter problems. Going through failure and finding solutions is a valuable experience, even more valuable than reading GAN’s paper from cover to cover.

directory

Since PyTorch Generation Vs. Network Programming is printed in color, I have included color catalog images. Please appreciate

Sample chapter appreciation