I, too, spent some time reading this book, and arrived at the conclusion that it's terribly written. From a mathematical perspective, the book does not define

I, too, spent some time reading this book, and arrived at the conclusion that it's terribly written. From a mathematical perspective, the book does not define hardly anything. Take, for example, the important chapter on Deep Feedforward Networks. They hem and haw saying things like "quintessential deep learning models" but never actually explain _what a deep feedforward network is_. There are no definitions: they give what amounts to preamble of a definition, and then move right on to using the notion, but don't actually give the definition itself. In the deep feedforward network example, the closest thing to a definition is, "A feedforward network defines a mapping y=f(x;theta)". That doesn't really say anything though. If you're asked to define "dog", you can say, "A dog has four legs", but how does the reader know that a cat isn't a dog? I find that for basic machine learning it's much better to just go through machine learning framework tutorials. For advanced learning, I'm not there yet so I can't speak from experience but I'll be shocked if this book fits well in any role except maybe last-resort reference material. If that.