Generative AI is the hottest topic in tech. This practical book teaches machine learning engineers and data scientists how to create impressive generative deep learning models from scratch using Tensorflow and Keras, including variational autoencoders (VAEs), generative adversarial networks (GANs), Transformers, normalizing flows, energy-based models, and denoising diffusion models. The book starts with the basics of deep learning and progresses to cutting-edge architectures. Through tips and tricks, readers can make their models learn more efficiently and become more creative.

• Discover how VAEs can change facial expressions in photos
• Train GANs to generate images based on your own dataset
• Build diffusion models to produce new varieties of flowers
• Train your own GPT for text generation
• Learn how large language models like ChatGPT are trained
• Explore state-of-the-art architectures such as StyleGAN 2 and Vision Transformer VQ-GAN
• Compose polyphonic music using Transformers and MuseGAN
• Understand how generative world models can solve reinforcement learning tasks
• Dive into multimodal models such as DALL.E 2, Imagen and Stable Diffusion for text-to-image generation

The book also explores the future of generative AI and how individuals and companies can proactively begin to leverage this remarkable new technology to create competitive advantage.

Objective and Approach

This book assumes no prior knowledge of generative AI. We will build up all of the key concepts from scratch in a way that is intuitive and easy to follow, so don’t worry if you have no experience with generative AI. You have come to the right place!

Rather than only covering the techniques that are currently in vogue, this book serves as a complete guide to generative modeling that covers a broad range of model families. There is no one technique that is objectively better or worse than any other—in fact, many state-of-the-art models now mix together ideas from across the broad spectrum of approaches to generative modeling. For this reason, it is important to keep abreast of developments across all areas of generative AI, rather than focusing on one particular kind of technique. One thing is certain: the field of generative AI is moving fast, and you never know where the next groundbreaking idea will come from!

With this in mind, the approach I will take is to show you how to train your own generative models on your own data, rather than relying on pre-trained off-the-shelf models. While there are now many impressive open source generative models that can be downloaded and run in a few lines of code, the aim of this book is to dig deeper into their architecture and design from first principles, so that you gain a complete understanding of how they work and can code up examples of each technique from scratch using Python and Keras.

In summary, this book can be thought of as a map of the current generative AI landscape that covers both theory and practical applications, including full working examples of key models from the literature. We will walk through the code for each step by step, with clear signposts that show how the code implements the theory underpinning each technique. This book can be read cover to cover or used as a reference book that you can dip into. Above all, I hope you find it a useful and enjoyable read!

Prerequisites

This book assumes that you have experience coding in Python. If you are not familiar with Python, the best place to start is through LearnPython.org. There are many free resources online that will allow you to develop enough Python knowledge to work with the examples in this book.

Also, since some of the models are described using mathematical notation, it will be useful to have a solid understanding of linear algebra (for example, matrix multiplication) and general probability theory. A useful resource is Deisenroth et al.’s book Mathematics for Machine Learning (Cambridge University Press), which is freely available.

The book assumes no prior knowledge of generative modeling (we will examine the key concepts in Chapter 1) or TensorFlow and Keras (these libraries will be introduced in Chapter 2).

Review

Generative Deep Learning is an accessible introduction to the deep learning toolkit for generative modeling. If you are a creative practitioner who loves to tinker with code and want to apply deep learning to your work, then this is the book for you.

David Ha

Head of Strategy, Stability AI

This book is becoming part of my life. On finding a copy in my living room I asked my son: "when did you get this?". He replied, "when you gave it to me", bemused by my senior moment. Going through various sections together, I came to regard Generative Deep Learning as the 'Gray's Anatomy' of Generative AI.

The author dissects the anatomy of Generative AI with an incredible clarity and reassuring authority. He offers a truly remarkable account of a fast-moving field, underwritten with pragmatic examples, engaging narratives and references that are so current, it reads like a living history.

Throughout his deconstructions, the author maintains a sense of wonder and excitement about the potential of Generative AI - especially evident in the book's compelling dénouement: having laid bare the technology, the author reminds us that we are at the dawn of a new age of intelligence. An age in which Generative AI holds a mirror up to our language, our art, our creativity; reflecting not just what we have created, but what we could create — what we can create — limited only by "your own imagination".

The central theme of generative models in artificial intelligence resonates deeply with me, because I see exactly the same themes emerging in the natural sciences; namely, a view of ourselves as generative models of our lived world. I suspect - in the next edition of this book — we will read about the confluence of artificial and natural intelligence. Until that time, I will keep this edition next to my Gray's Anatomy, and other treasures on my bookshelf.

Karl Friston, FRS

Professor of Neuroscience, University College London.

Generative AI is reshaping countless industries and powering a new generation of creative tools. This book is the perfect way to get going with generative modeling and start building with this revolutionary technology yourself.

Ed Newton-Rex

VP Audio at Stability AI and composer

An excellent book that dives right into all of the major techniques behind state-of-the-art generative deep learning. You'll find intuitive explanations and clever analogies -- backed by didactic, highly readable code examples. An exciting exploration of one of the most fascinating domains in AI! 

Francois Chollet, Creator of Keras

Generative AI is the next revolutionary step in AI technology that will have a massive impact on the world. This book provides a great introduction to this field and its incredible potential and potential risks.

Connor Leahy, CEO at Conjecture and Co-Founder of EleutherAI

About the Author

David Foster is a Founding Partner of ADSP, a consultancy delivering bespoke data science and AI solutions. He holds an MA in Mathematics from Trinity College, Cambridge and an MSc in Operational Research from the University of Warwick.

Through ADSP, David leads the delivery of high-profile data science and AI projects across the public and private sectors. He has won several international machine-learning competitions and is a faculty member of the Machine Learning Institute. He has given talks internationally on topics related to the application of cutting-edge data science and AI within industry and academia.