Click above “ Program life ”, choice “ Official account ”

First time focus on procedural apes ( Yuan ) Stories around us

Comparable TensorFlow Deep learning framework of !

actual combat Kaggle Classic events in the competition

GAN Generate animation Avatar ,AI Filter ,RNN Writing poetry , Image description task

2016 Year belongs to TensorFlow A year of , Thanks to Google's promotion ,TensorFlow Dominated the headlines of major media .2017 Beginning of ,PyTorch The emergence of the new technology has attracted great attention of researchers ,PyTorch Simple and elegant design , Unified and easy-to-use interface , The speed of chasing the wind and the flexibility of changing are impressive .

As a door 2017 Deep learning framework just released in , Researchers have limited access to learning materials , Chinese materials are even less . Long term concern of the author PyTorch development , Often help in forums PyTorch Novice problem solving , Use in scientific research at ordinary times PyTorch Conduct research in all aspects , With rich experience . See the domestic users PyTorch Very interested , There is an urgent need for a comprehensive explanation PyTorch 's books , So this book was born .

《 Deep learning framework PyTorch: Introduction and Practice 》

Main contents of this book

This book from multidimensional array Tensor start , Step by step to guide readers to understand PyTorch Basic knowledge of all aspects . Combining basic knowledge with cutting-edge research , Lead readers to complete several classic and interesting in-depth learning projects from scratch , include GAN Generate animation Avatar ,AI Filter ,AI Writing poems, etc .《 Deep learning framework PyTorch: Introduction and Practice 》 There is no simple and mechanical introduction to the use of each function interface , It's about trying to categorize , Introduce the reader step by step PyTorch Knowledge of , I hope the readers are right PyTorch Have a complete understanding .

The content of this book goes from shallow to deep , Whether it's a beginner of deep learning , First contact PyTorch Of researchers , Can be mastered quickly in the process of learning this book PyTorch. Even if there is a certain PyTorch Users with experience , You can also get the right answers from this book PyTorch Different understanding .

About the author

economic expert :Python programmer ,Linux Enthusiasts and PyTorch Source contributor . Main research fields include computer vision and machine learning .“2017 Zhihu watch mountain cup machine learning challenge ” the first prize ,“2017 Tianchi medical AI Competition ” Eighth place .
Keen to promote PyTorch, With rich experience , Active in PyTorch Forum and Zhihu related sections .

Structure of the book

The book is divided into two parts : The first 2~5 Chapter introduction PyTorch Basic knowledge of .

√ The first 2 Chapter introduction PyTorch Installation and configuration of learning environment . At the same time, it is introduced in the most general way PyTorch Main content of , Let readers PyTorch Have a general impression .

The first 3 Chapter introduction PyTorch Multi dimensional array Tensor And dynamic graph autograd/Variable Use of , With examples , Let readers use them separately Tensor and autograd Realize linear regression , Compare their differences . This chapter also Tensor Bottom design of , as well as autograd The principle of , Give readers a more comprehensive and specific explanation .

The first 4 Chapter introduction PyTorch Central neural network module nn Basic usage of , At the same time, the neural network “ layer ”,“ loss function ”,“ optimizer ” etc. , At last, it can't be used to guide readers 50 The code of line was built to win ImageNet Champion ResNet.

√ The first 5 Chapter introduction PyTorch Data loading in ,GPU Relevant tools such as acceleration and visualization .

The first 6~10 Chapter mainly introduces actual combat cases .

The first 6 Chapter is a connecting chapter , The goal is not to teach readers new functions , New knowledge , It's a combination Kaggle A classic game in , A simple image two classification problem in deep learning . In the process of implementation , Before leading readers to review 5 Chapter knowledge , Code specification is proposed to organize program and code reasonably , Make the program more readable , Maintainable . The first 6 The chapter also introduces PyTorch How to proceed in debug.

√ The first 7 Chapter 3 explains the most popular generative network for readers (GAN), Lead the reader to realize a cartoon image generator from scratch , Able to use GAN Generate animation avatars with changeable styles .

√ The first 8 The chapter explains the knowledge of style transfer for readers , And lead readers to realize style migration network , Turn your photos into “ high-end, sophisticated and classy ” Famous paintings of .

The first 9 Chapter 2 introduces some basic knowledge of natural language processing , And explain CharRNN Principle of . And then use it to collect tens of thousands of Tang poems , Train a small program that can write poems automatically . This small program can control the format and artistic conception of poetry , Can also generate Tibetan Poems .

√ The first 10 Chapter 2 introduces the task of image description , And with the latest AI Challenger For example, the data of the competition , Lead the reader to implement a simple image description of the small program .

The first 1 Chapter and Chapter 11 Chapter is the first and last chapter of the book , The first 1 Chapter introduction PyTorch Advantages of , And the comparison with other frameworks on the market . The first 11 Chapter is a summary of the book , And right PyTorch Thinking about the shortcomings , At the same time, some suggestions are put forward for the future study of readers .

Readability group

To learn this book, you need to have the following basic knowledge :

√ understand Python Basic grammar of , Basic Python usage method .

√ Have a deep learning foundation , Understanding back propagation , Convolution neural network and other basic knowledge , But it doesn't require a deep understanding .

√ With gradient , High school mathematics basic knowledge such as derivative .

The following knowledge is not required , But it's better to know :

√ numpy Use of .

√ The basic process of deep learning or the use of other deep learning frameworks .

Scan QR code purchase 《 Deep learning framework PyTorch: Introduction and Practice 》

Last period “ Letter of recommendation ” Announcement of award list

award-winning 3 Famous students

Please leave your name in the comments , Contact information and address

After reading this , What are your views and ideas , Welcome to leave a message and discuss with us .


Click on the picture get Previous content