[disgruntledphd2]

I enjoyed the book, but have a background in this area.

I was amused by the suggestion that computer science undergrads could handle the book, as clearly the authors and I have met very different computer science undergrads.

[mlthoughts2018]

The Courville, Goodfellow and Bengio book is definitely suitable for undergraduates. In my current job, we often have new junior level (bachelor’s grads) ML hires work through that book and present chapters in the team reading group. In my experience both as a TA in my PhD program and in industry, that book is fairly easy to read through for anyone with solid understanding of linear algebra and vector calculus, which are freshman / sophomore level college math courses.

[zer0sugar]

Here's a photo of a random page.

https://i.imgur.com/vv1CRLv.jpg

You can trust me when I say the entire book is about as unreadable as that and often worse. I'm not afraid of math either. But the book certainly is not teaching anyone anything.

[random314]

I am astounded by how you continue to insist that the book doesn't teach anyone anything, when I have already stated that I learned something from it! And of course the book has equations in it. What did you expect?

[zer0sugar]

What did you learn from it?

None of my math books are as obtuse as it is. The equations are presented on their own without explanations. On that page alone they're using quite a bit of mathematical notation that I, at least, have never seen before and I suspect it's largely unnecessary.

What did I expect? I expected a book that explained the concepts in plain english as well as mathematically. I expected the authors to be mature enough not to heavily decorate every single equation with as much mathematical notation as possible. Sort of like how bad coders make their code hard to read. That's the vibe I'm getting from the book.

[random314]

I learnt eigen decomposition, Hessians, PCA, backpropagation, CNN, dropout, maxpooling etc.

The page you linked above is the derivation of PCA using linear algebra.

First part derives the encoding matrix from the decoding matrix. 2nd part derives the encoding matrix by minimizing the L2 norm.

If you find the math too heavy, you should take Andrew ngs course at Coursera (not his Stanford lectures, which follow a pattern similar to this book). Or pick up any book targeting programmers, machine learning for hackers etc.

[disgruntledphd2]

Cool, I'm glad it's been working out for you. Don't get me wrong, I enjoyed that book, even the start which wasn't focused so much on deep learning specifically.

I just don't know many computer science undergrads who'd have the background to make that book useful, as the presentation leans towards the terse.