For anyone who is serious about Machine Learning, Linear Algebra background is a must. I was asked several times by several people about a good introductory textbook in Linear Algebra, and every single time I struggled to recall the name of the textbook I used.
But here it is, I somewhat magically found it very recently
It is super concise and greatly educational. Digesting the whole 300 pages will prepare you more than enough for pretty much any Linear Algebra stuff you will find in Machine Learning.
I learned Linear Algebra in undergraduate (with pretty decent professors I have to say), but it is in this book where my moment of enlightenment happened. Very early in the book, it presents how we should see the matrix multiplication in a slightly yet radically different view: the multiplication should not be interpreted as multiplying every row of with , but actually it is using elements of to create a linear combination of the columns in . This is the fundamental of all the reasoning in Linear Algebra, because with this view, belong to the column space of (a.k.a the range of A). Every Linear Algebra reasoning becomes super clean with this interpretation. Every matrix decomposition you find in Machine Learning (SVD, Cholesky, LU…) suddenly becomes all easy.
In other words, this is highly recommended for any newbie in Machine Learning. I believe you will enjoy it as much as I did.