Strang's “Linear Algebra and Learning from Data” is printed and available
math.mit.edu
math.mit.edu
Always love it when a professor can bring in some comic relief in the midst of a very heavy math topic. The students seem to enjoy it. I am self-teaching myself background math for preparing me to the likes of PRML-Bishop, and I wholeheartedly recommend his Linear algebra course available on MIT Courseware[1] coupled with his book[2]
[0] https://www.youtube.com/watch?v=amv58LCqCMI [1] https://ocw.mit.edu/courses/mathematics/18-06-linear-algebra... [2] https://www.amazon.com/Introduction-Linear-Algebra-Gilbert-S...
I have to be honest, I never have understood why everyone loves his texts so much. I always felt that they were lacking in practical applications (even the Linear Algebra with Applications version) and that a lot of the ideas were not immediately clear when presented in his writing style. It often seemed that ideas and definitions would be thrown out with not much connection to why they were interesting. (Although I suppose this can be said of many math courses).
I suppose the thing is, with this stuff, there are lots of really interesting applications. I think I would have preferred something that started with the applications and then worked backwards. Maybe this book is an improvement? I'd like to check it out.
The exercises also didn't seem particularly enlightening, I just remember doing lots of rote things like LU factorization by doing Gauss elimination by hand.
I will note, however, that I really like the fact that the new book has a chapter on compressed sensing. I feel like that was the thing that was always lacking when I was in school. It was like "ok, I get the idea of new bases, now what are some cool ones and how do they tie together." Occasionally someone would throw me a paper on compressed sensing or PCA/SVD or spectral decomposition, but I always felt like I wanted to see a class/book that tied them all together.
[0] https://youtu.be/gGYcSjrqbjc
[1] https://ocw.mit.edu/courses/mathematics/18-06sc-linear-algeb...
I had to rapidly teach myself Linear Algebra to get through some machine learning graduate courses, and Strang's book and video lectures saved me. He has a very special talent for making complex material understandable without dumbing it down. He's up there with Norvig and Feynman for me in terms of teaching talent.
That man with a piece of chalk and a big chalkboard beats any PowerPoint presentation or video presentation using animation and whatever other technological helps you can think of.
This is a textbook to help readers understand the steps that lead to deep learning. Linear algebra comes first especially singular values, least squares, and matrix factorizations. Often the goal is a low rank approximation A = CR (column-row) to a large matrix of data to see its most important part. This uses the full array of applied linear algebra, including randomization for very large matrices. Then deep learning creates a large-scale optimization problem for the weights solved by gradient descent or better stochastic gradient descent. Finally, the book develops the architectures of fully connected neural nets and of Convolutional Neural Nets (CNNs) to find patterns in data. Audience: This book is for anyone who wants to learn how data is reduced and interpreted by and understand matrix methods. Based on the second linear algebra course taught by Professor Strang, whose lectures on the training data are widely known, it starts from scratch (the four fundamental subspaces) and is fully accessible without the first text.
In addition, ~40% of the book is devoted to basic probability and statistics, optimization, and neural networks.
This isn't a valid "critical perspective"