Math for Deep Learning: What You Need to Know to Understand Neural Networks

Dr kneusel does an amazing job breaking down neural networks into the fundamental components as well as statistical frameworks . Highly recommend for anyone looking for an intro to or refresher on the math required for ML. It will require some studying though if one does not have a solid maths background .

My background: PhD -- MIS, ABD -- Statistics; MBA; most of MS EE; BS ChEI believe the ideal audience for this book are those that want to be a data scientist -- probably coming from a somewhat technical background, but not so well-grounded in math. Perhaps taking one of the many boot camps extant. For that audience, this is an excellent generalist book but does require motivation and/or previous math skills.Generally well written, it weaves together math skills and relating how this math is important to understanding many of the core concepts of machine learning.It is light enough for bright, motivated folks to gain a critical understanding of much of the math necessary to understand ML more deeply and intuitively. Heavy enough to be more than a fluffy treatment that would have the persistence of carbonation.The chief improvement would be a worked exercise set, some covering the math; the rest applying the math in simple ML exercises. But, of course that would have been a much longer book to write.As it is, a low 5-star review seems well-earned.

“There should be no such thing as boring mathematics.” This book just nailed it big time. One of the best books on applied math in action. Not too intimidating math but sufficient enough to build Deep Learning math intuition end-to-end on serious level.Definitely not your ”leisurely weekend reading” but intriguing enough to invest some time to work through it. The author managed to keep the readers engaged, not to scare them off, at the same time was pragmatic about need to develop decent applied math chops.It is always hard to balance “why’s” (mostly proteins, focus on building blocks in cohesive way) with “how’s” (lots of empty calories, endless trial and error with hyperparameters etc). This book is a good “dietary” recipe for getting one’s feet wet in Deep Learning the right way :-)Great job Dr. Kneusel!

I've skimmed the entire book and have worked through several chapters in detail. My first reaction is this book covers about 3-4 years of undergraduate math with some graduate school math topics.My CS degree is 30+ years old, but my daughter just finished her CS/Math degree and the undergraduate math requirements have changed very little. My old-school CS degree covered only about 80% of the topics in this book.If you've taken these classes then this is book provides an awesome review put in the perspective of how to use python to help with these problems. If you haven't taken at least statistics, calculus, and linear algebra, you may find that you need books that introduce these topics in a more beginner-friendly way. And then use this book to provide ML context to the math. There are many good intro books and youtube videos that can help start with these topics. An example of a beginner-friendly statistics book might be "The Cartoon Guide to Statistics" which is fantastic. Good videos would be anything by 3blue1brown.I'm reading the Kindle version of the book. The formatting seems quite good for a Kindle book. Kudos to the author for not treating Kindle as an afterthought.I'm finding this to be a very useful book. The title did warn that it's a mathy book. I think to get a good feel of ML right now, this math background is probably necessary. And this book seems quite good.

This is a Math Book, but does not have anything applicable to Deep Learning!The author spends three pages on the Monty Hall Problem (For those who don't know what the Monty Hall problem is, it was a TV show thirty years ago, called, "Let's Make a Deal".) What has the Monty Hall problem to do with Deep Learning.It would have been useful if the author had taken a single example, say, the recognition of hand-written digits (the famous MNIST), and worked from the beginning to the end, including the back propagation, and introducing the necessary Math as he proceeds, it would have been immensely useful.I can think of at least one use for this book: Use it as a paper weight.

I appreciare the author effort to select necessary maths for Deep learning. Most of maths books for deep learning are too vague for engineers or general data analyst. However, the math listed in this book are a bit over shallow that understand all of these are far not enough to advance deep learning. I would say this book is just a light reading for a relax weekend, but nothing more.