A Roundup Review of the Latest Deep Learning Books

For years, nobody wanted to read about AI. It was a backwater of research, solving toy problems while crashing and burning on real world challenges.

But then came a flurry of seemingly overnight breakthroughs:

Image recognition; self driving cars; Alpha Go.

The old algorithms, first dreamed up in the 80s and 90s, suddenly started to work, fueled by the power of massively parallel chips and big data sets.

Now researchers are racing to fill the gaps on your bookshelf. For the last year they’ve pounded the keys, writing as fast as their fingers could carry them, hoping to beat each other to market with a fantastic book that will feed the growing hunger for deep learning knowledge.

The first of those books are already hitting the shelves. More are coming this summer and early next year. I’ve had a chance to read through a number of their early drafts and I’m super excited to see some of them finally go live.

So let’s dive in and take a look at a roundup of the best and worst books to start learning AI or advance your burgeoning understanding of the art.

Balancing Act

My father always said “balance is best in all things.”

I live by these words.

OK, you got me, I occasionally ignore them on the weekend or in Vegas.

That said, they mirror my learning style. I favor a balance of theory and practicality. Give me the background behind an idea in clear language and then let me experiment with practical examples.

But don’t give me too much theory. I want to get my hands dirty when I learn. If a book veers too far into the abstract, or drops a bunch of examples on me with no context I quickly lose interest.

That said, respect your own learning style. Know how you learn best. If you love a book with pages and pages of detailed theory, then just pretend you’re reading my review with an inverse filter. Go buy all the books I hated and ignore the ones I loved!

But if you’re the type of person who loves a balanced teaching style, then this is your list.

Run, don’t walk, over to Amazon and one-click these deep learning wonders.

The Contenders

Our first challenger is Ian Goodfellow’s Deep Learning. No list would be complete without this Google Brain and Open AI research star’s work. Already, some people consider it the bible of deep learning, the only book to bring together decades of research in a single magnificent tome.

But unless you have a serious background in mathematics, this is absolutely not the place to start. It will only frustrate you.

Not only is it filled with page after page of mego inducing equations, it’s also written in dust dry, text book prose. While you have to respect Goodfellow’s brilliant mind and his desire to cram as much as possible between two covers, it doesn’t exactly make for riveting reading.

The reason is simple. There’s a difference knowing something and teaching it. I suspect this book will end up in the hands of many aspiring students signing up for deep learning classes in college for the first time next year.

And it will drive many of them to quit.

If you’re looking to master deep learning, after years of study, this might be the book for you. There is no more comprehensive guide on the planet. It’s a book that assumes a lot of domain knowledge already. But if you’re just starting out or coming to AI as a professional programmer for the first time, this densely packed text book will leave you wanting.

Next up with have Hands-On Learning with Scikit-Learn and Tensorflow hot off the press! While this book is also packed with equations, it’s still incredibly readable. In fact, it’s downright amazing. I cannot recommend it highly enough. The Learning AI if You Suck at Math article series is deeply in this book’s debt, especially article five on image recognition with convolutional neural nets and article seven on natural language processing. Author Aurélien Géron has a way of making complex topics accessible to a general audience that I try to mirror in my own style.

To my mind, this is hands down the most perfect balance between beautifully explained examples and workable code. I read early drafts of it on Safari Books Online. Even with many of the parts unfinished, and with the website turning parts of some equations into unintelligible mush, it stood out as a fantastic read that really leveled up my understanding.

The final release version seriously improves on the original drafts. Like all good rewrites, it makes the book as a whole better. The ideas and examples are crisper and better explained. It’s also organized with a more natural flow that really brings the topics across with enough clarity to experiment and enough depth to go back and learn more on a second rereading. You may find yourself skipping most of the equations on the first go around, only to go back and study them more closely.

Number three is Deep Learning with Python by Keras creator Francois Chollet. This is one of those books that just can’t come out fast enough. Pre-order this one immediately. I read the first three chapters via Manning Press’ MEAP program, aka Manning Early Access. While it might seem a little early to give it my top recommendation based on three chapters alone, I do it without reservation. It’s that good.

Just as Chollet has a knack for simplifying complex concepts with code in Keras, he brings the same engaging and beautifully readable style to his writing. The book makes it easy to understand even the most challenging aspects of AI and deep learning. I didn’t understand a damn thing about tensors until I read this book but he helped me break through the fog and see them for exactly what they are: buckets for numbers. And as you might expect, the book also has some excellent examples, considering Chollet’s Github is filled with some of the most forked AI code on GitHub.

I can only see this book getting better as it gets closer and closer to release. Grab it now to support him. If you can get a copy on MEAP, do that as well. You can even improve the book in true open source style by providing feedback to the man himself!

Deep Learning: A Practitioner’s Approach is book number four on our list. It focuses on the java framework DL4J. While so much of the research in AI is done in Python, it’s incredibly likely that we’ll see a lot of that work shift to Java as more and more enterprises embrace machine learning. Java remains the number one language in big companies. It’s portable, reusable and has an army of classically trained programmers who know it better than anything else.

One of the co-authors of the book, Josh Patterson, is presenting with me on AI at Red Hat Summit, coming up in the first week of May. I’ve had a chance to read through the nearly ready for print version of the book and it’s fantastic. To be clear, this is a first time learner’s book on deep learning. If you already have some background and you just want to explore DL on Java, you’ll want to skip ahead to the examples. But if you have little to no DL experience and a strong grounding in Java, this is the book you’ll read cover to cover. Chapter Four: Major Architectures of Deep Learning is a particular standout. It offers a terrific round up of the key architectures that help you solve practical problems today.

While I am not a Java programmer by any stretch of the imagination, I’ve shared it with several of my colleagues who live and breathe the language and they love it. I found the examples and overall organization of the book to be near perfect as an introduction to DL. Expect it to release late summer.

Last up we have the TensorFlow Machine Learning Cookbook. This book suffers a little from typos, including in the code, but overall it provides a number of decent to good examples on various topics like natural language processing.

That said, I would not consider buying this book in isolation.

Like any cookbook it leaves much of the deeper explanation to other books and focuses on the code almost exclusively. If you don’t already know the ins and outs of a convolutional neural network, you’ll be struggling to figure out a lot of concepts that are simply left unsaid. If you want to buy this book after you’ve read one of the others and experimented with all their examples, this one could help you with additional practice and exercises. Just don’t start with it.

And That’s That

There are a few more books coming down the pipe, as well as a few more already out now, but I haven’t had a chance to check them out just yet. Once I’ve had a moment to look at some of the others in the coming months, I’ll follow up with another roundup review. But for now, you can’t go wrong with some of the excellent choices right here.

So what are you waiting for?

Get started.

The power of deep learning is in your hands now!

############################################

If you enjoyed this tutorial, I’d love it if you could hit the little heart to recommend it to others. After that please feel free email the article off to a friend! Thanks much.

###########################################

A bit about me: I’m an author, engineer and serial entrepreneur. During the last two decades, I’ve covered a broad range of tech from Linux to virtualization and containers.

You can check out my latest novel, an epic Chinese sci-fi civil war saga where China throws off the chains of communism and becomes the world’s first direct democracy, running a highly advanced, artificially intelligent decentralized app platform with no leaders.

You can get a FREE copy of my first novel, The Scorpion Game, when you join my Readers Group. Readers have called it “the first serious competition to Neuromancer” and“Detective noir meets Johnny Mnemonic.”

Lastly, you can join my private Facebook group, the Nanopunk Posthuman Assassins, where we discuss all things tech, sci-fi, fantasy and more.

############################################

I occasionally make coin from the links in my articles but I only recommend things that I OWN, USE and LOVE. Check my full policy here.

############################################

Thanks for reading!