PyTorch Deep Learning Hands-On: Build CNNs, RNNs, GANs, reinforcement learning, and more, quickly and easily

by Sherin Thomas, Sudhanshu Passi

Developing image analysis apps, GAN-based networks, reinforcement learning algorithms and text engineering routines with Deep Learning PyTorch applications

Key Features The first book-length introduction to PyTorch Covers the whole range of possible applications that can be written on PyTorch Focuses on the APIs, and treats algorithms as secondary Book Description

Deep Learning is probably the fastest-growing, but also the most complex area of applied computing today. There are two major frameworks dominating the Deep Learning API landscape – Google’s TensorFlow, and Facebook’s PyTorch. Deriving from the open source Torch framework written in Lua, it was under the leadership of AI guru Yann LeCun that Pytorch developed into a major alternative.

PyTorch uses autodifferentiation to make it possible for developers to introduce new behaviors into their neural networks, without having to restart their networks. This is possibly the most important innovation for major machine and deep learning frameworks implemented in Pytorch. Also, PyTorch threads can run on CPUs as well as GPUs, providing major efficiency gains in the process.

This book shows us how to make the simplicity and power of Pytorch work for a Python developer. The first application we learn about is how how to process images using CNNs, but new algorithms like GANs and and natural language processing algorithms are introduced as well. The book ends with a chapter on reinforcement learning and how put PyTorch application into production

What you will learn Processing, improving and recognizing image features Finding, interpreting and deriving insights from unstructured textual data Learning several varieties of General Adversarial Networks (GANs) Apply PyTorch implementations of reinforcement learning algorithms Put PyTorch projects through a production cycle Who This Book Is For

Fluency in Python is assumed. Basic deep learning approaches should be familiar to the reader. This book is meant to be an introduction to PyTorch, and tries to show the breadth of applications PyTorch can be put to.

About the Author

After doing his computer science degree in Kerala, Sherin Thomas became a developer and AI expert for various Indian companies. A strong interest in open source AI led Sherin to start writing and presenting at conferences, culminating in this book on PyTorch.

250 pages, Paperback

Published April 30, 2019

Reviews

[Arthur · Rating 5 out of 5]

After I already picked a copy of the book I suddenly realized it is only 200 pages plus so what would I learn?
Quite a bit. Despite being relatively succinct at first glance the author managed to actually squeeze in material beyond the basics. Not to server as a spoiler, this book has helped me to advance on my own terms without spending too much time in the book. The learning process basically starts right away, thus the Hands-On in the title.
Good points: most of the tools are the common ones, Python, Ubunty. With a minor exception being Flask that I really was never exposed to. But frankly I find it not intimidating as some other Web Frameworks I prefer not to mention. One of very few books that explain things visually as Neural Network kinds. Overall a good coverage of the most typically found in use NNs.
What I found a bit difficult to understand - formulas.
What could be improved: code annotations are too sparse (not everywhere).
Still it is a 5/5 for me.