Hi Everyone ! I am desperately looking for the right textbook / course that will fit my needs. I am looking for a good linear algebra text that will find application in statistics and/or ML. Basically that's what I did : Linear algebra : - MIT courses ( not really deep, expose basic concepts without too much digging) - Read Linear algebra done right ( Loved this book, cause it is theoretical and easily understandable ) - Read a large variety of Linear algebra book but all of them expose theorem, proof and move on
Machine learning : - NG courses on coursera and standford ( unfortunately not really deep, for example at standford he write some linear algebra formula without proving them which is a fundamental mistake at this point, he suppose student are going to figure the formulas out by themselves. ) But anyway the content is not really deep, some concept fall out of nowhere. - Statistical learning. That's a good one but once again too many " The proof goes beyond of this book " So frustrating ...
Finally I took the Linear dynamical system course from standford. I really liked it, it introduced some major concept in a very friendly way, BUT STILL no really deep mathematical explanations.
I guess what I am looking for is formulas and concept used in statistics and machine learning which comes with the full Linear algebra theory. 'With all my readings I managed to get down the linear regression problem from bottom to top.
I was wondering what are your recommendations on that topic. I'm starting to get really tired of buying book that I stop to read half way, and my constant looking around.
I know books and courses depends higly on the reader profile, so this is my profile : I absolutely love all mathematical concept as long as you introduce them with enough explanations, no " the proof is beyond this book" or " I'm not going to show you a proof of this ". I am more interested about linear algebra than calculus, as LA if understood well generalize a lot. A good book in linear algebra that explain even the basic concept ( eigendecomposition, SVD, vector space subspace ) is good to take as long as theorem are not thrown away in your face. My perfect book or course will be a statistical / ML book that will go into awfully long Linear algebra details.
I'm starting to take the standford course Convex optimization but i'm not sure about it. Since courses take a long time to process ( and in my case really long time cause I try to prove everything on paper ), I would like your input on it too.
Thank you all !
[link][6 comments]