I think I got to chapter 6 of deep learning(deeplearningbook.org) before I ran out of steam. I went non-traditional as a Chem then physics-pre med major before grad school in comp sci so my linear algebra is piecemeal from quantum physics and math methods in physics. I really wish I took linear algebra before going through my cs grad school, but now I"m almost tempted to do Gilbert Strang's course on OCW to brush up on it. Linear algebra is the key to this stuff--well some tensor analysis as well, which, in my case, Dr. Houston snuck into math methods using Einstein notation. Yes, I did the Tensor Flow examples with the MNIST data, which was neat. I'm going to try the Coursera course as suggested by this article. I have links to Strang's linear algebra as well as a Drexel University course on Tensors on my page (
https://hackereducation.wordpress.com/hacker-education-conte...)