There is a deluge of tutorials claiming to teach you deep learning in 15 minutes or less, promising to keep your career safe from the robots while skyrocketing your salary. This type of learning serves as a good introduction, but is unlikely to teach you what is actually happening when a computer program learns to recognize a face, or generate text. I learned of ‘Grokking Deep Learning’ from this lecture on the state of the art of deep learning, and I found this book to be the opposite approach; a deep dive into deep learning.
The book starts with almost no assumptions of prior deep learning knowledge. All you need is basic Python and high school level math. Starting from this level, the book layers concepts on top of each other until you understand what even very complicated networks are doing when they learn. Trask’s method of teaching in this book shows you precisely what happens, in terms of numbers, when a network learns to solve a task. His code is very concise, leading to quick models where you can verify yourself that the computer is learning, and see how that is happening.
I would highly recommend this book. Learning new concepts in computer science or mathematics is often difficult, but Trask progresses inch by inch until each concept is clear. Thanks to resources like these, I can progress my own career in machine learning, and hopefully one day propagate my own knowledge on the subject. In the meantime, I have a lot more deep learning to do.
You can see the book here
Andrew’s community OpenMined, where he explores AI privacy and (eventually, probably) advances the field of cancer research by making health data both available and private.
My Github, where I reproduce Andrew’s results with my comments, and do other deep learning experiments.
Disclosure: These words are my own. I have no affiliation with Andrew Trask or OpenMined.