Grokking Deep Learning

This is a great introduction to deep learning. It throws light on the fundamental concepts of deep learning, and helps with practical code snippets. It gives some hands on experience to the reader to build simple models on realistic use cases. It provides progressive, code-first approach to reinforce the advanced concepts being discussed.

The author has done an astounding job at distilling the essence and complexities of the topic into a format that is almost child’s play. At first this book for me was a supplementary read whilst I did the fast.ai course, but since has become the main focus and resource for learning. Thank you Andrew for bringing this topic to the layman and propelling me Into this field.

Awesome book on Deep Learning. It builds up concepts in a lucid manner. The coverage is comprehensive and logical. Must for any one who is interested in Deep Learning.

Despite having a mathematical background (and having grokked many concepts of DL) I found this book insightful. On some shortcomings: I agree with some of the other reviewers that, having mastered the maths of DL, this approach might rather confuse. Maybe math is the simplest language in this regard. Personally I found a few points more distracting than this, specifically the backpropagation for the NN for sigmoids in chapter 11 (an approximation is used which is never explained) and the batch gradient descent in chapter 8 (just wrong). A minor point: Personally I prefer PEP8 coding standard, the code is also a bit unique and sometimes hard to read. On its strengths: This book invites to deal with DL deeply, try out for yourself, to learn from your mistakes (and maybe deliberately from shortcomings in the explanations). This is sometimes painful, yet this is the way you learn. Specifically the strongest points are the step-by-step explanations (e.g. backprop) where each step is executed "on paper". I very much liked the part on Autograd. Having Worked with PyTorch and TF for years I took many things for granted which, surprisingly, I really never understood. Building a framework step by step from ground up helped me a lot. An other great point are the many tips throughout the book (sometimes well hidden), which are also worded so that they invite deeper understanding (and more research into the topic). Overall (for me) the strong points outweigh the shortcomings by a wide margin.

This book came highly recommended to me by a colleague at work. I and others at work have agreed that it's a great primer to learn how machine learning works and how to build your first model. It walks through every step and detail with examples, code, and visuals to bring it all together.