At all times you should be reading a book which is too hard for you to read.
These books either have or still do fulfil that criteria for me.
They’re the foundation upon which I’ve built my knowledge of machine learning and data science. I’ll continue to read and reread these for years to come.
And if you’re learning machine learning or data science, they’re worth your time.
If I’m learning something new, I’ll usually find a good book on the topic and read it end-to-end. I’ll then follow up on the parts that stick. These books have plenty of parts which have stuck.
Books are listed in order of approachability (roughly), if you have 0 experience in machine learning or data science, start from the top, if you’ve got Python and math down pat, go from the bottom.
Machine Learning for Humans by Vishal Maini and Samer Sabri
This book started as a series on Medium. The authors wanted to explain all they knew about machine learning in a readable and approachable way. And they’ve done just that.
If you want a zero-to-one resource you can use to build an understanding of some of the most important machine learning concepts, but you haven’t encountered machine learning before, this book is for you. Even if you’re already a machine learning practitioner, this book is worth reading. It’ll give you inspiration for sharing your work in way which is approachable for others.
Python for Data Analysis by Wes McKinney
Start learning data science or machine learning and you’re going to be using Pandas (a Python library for data analysis). The best thing about this book is it’s written by the creator of Pandas so you know you’re learning from the best.
As a machine learning engineer, I spent most of my time using Pandas to manipulate data to get it ready for machine learning models.
This book will show you how to use Pandas to analyse your data, clean it, change it and most of all, use it for data science and machine learning.
As a data scientist or machine learning practitioner, you can never have enough Pandas knowledge.
Hands-on Machine Learning with Scikit-Learn and TensorFlow by Aurelien Geron
If you’re getting into machine learning and you want a one-stop practical resource, this is it. It’ll take you through two powerful machine learning libraries, Scikit-Learn and TenorFlow and teach you machine learning concepts through coded examples.
Each concept has code to go along with it. So you could read this book, get an understanding of what machine learning is capable of, then adjust the code examples to your own problems.
Grokking Deep Learning by Andrew Trask
I started learning deep learning via Udacity’s Deep Learning Nanodegree. Andrew Trask was of one of the teachers. He’s now a researcher at DeepMind.
Back then, there was only a few of chapters released. I sat on my couch flicking through page by page, learning how to build a neural network from scratch with NumPy (a Python numerical library).
I was hooked on the descriptive analogies he used to describe machine learning concepts.
“Deep learning hyperparameters can be tuned like the dials on your oven.”
I devoured each new chapter as it came out.
But now you don’t have to wait, the full book is ready.
This book is a chance to learn deep learning from the ground-up and with hands-on examples from one of the best practitioners in the field.
The 100-Page Machine Learning Book by Andriy Burkov
The start here and continue here of machine learning. That’s what I called it my book review. After reading Machine Learning for Humans, if you’re hungry to get deeper on what makes machine learning algorithms tick, this is the book for you.
My favourite part is it covers problems in machine learning and gives you solutions, as well as the rational behind those solutions. All within 100-pages.
You could read this in a day if you want. But you don’t need to. Take your time. Learning anything new takes time. Especially machine learning.
If the 100-pages aren’t enough, there’s QR codes scattered throughout with extra-curriculum curated by the author.
The Deep Learning Book by Ian Goodfellow, Yoshua Bengio and Aaron Courville
This is the newest edition to my collection. I bought the hard copy. It’s the book which fulfils the criteria at the start of the article.
I’m most excited for the math sections at the start. I’ve been a code-first learner. Hence the order of these books. But deep learning and machine learning are based on applied math. The code and frameworks might change over time but the math doesn’t change. Linear algebra is always going to be linear algebra.
The Deep Learning Book is written by three titans of the deep learning world. Goodfellow is the inventor of GANs, Bengio is one of the original discovers of deep learning and Courville’s academic works have been cited nearly 50,000 times.
This book dives deep on all of the deep learning concepts you should know about (not a pun).
Remember, machine learning is broad. Use these books as a foundation to base your knowledge on and improve it by getting your hands dirty.
Knowledge which isn’t applied is wasted. There’s no better way to learn than to make mistakes.