|
I assume you're looking to do an ML PhD, so you probably don't need to learn up on number theory. Mathematics is a huge topic, so you need to condense what you're going to study down to a couple of courses. Also, ML doesn't use that much actual maths, so you don't need to know proofs, real analysis, algebra, differential equations, etc. Though doing a a decent proofs course is always helpful everywhere. I would start with linear algebra. This post [1] links to a great course. Then you might want to do a course on multivariate calculus and a course on optimization but it's not really required. Second, I would do this [2] course on machine learning. After these two courses, which would probably take you 3 to 4 months if you have a job, then you'll have a better idea of what math courses to take. Final note, when I say do a course, I mean watch a lecture, take notes like you would do in a real class, go to the course website, go through the lecture notes, do the homework set associated with the lecture, and then go to the next lecture. Just watching the lecture, especially for a math lecture, especially of you're not familiar with the subject, is not as helpful. [1] https://news.ycombinator.com/item?id=19812266
[2] https://www.youtube.com/playlist?list=PLA89DCFA6ADACE599 |