What area of ​​mathematics do you need to master in order to understand deep learning at a deep level?

Question

Mathematics

Actually the question is in the title. I'm tired of just using and implementing invented algorithms in machine learning. I already want to move more towards the mathematics of this area. What is worth studying/reading?

Answer 1

You can start with this:
And then from above:
Efimov, Rozendorn - linear algebra and multidimensional geometry
Then polish:
And then immediately:
This is all about solving problems with ordinary mathematics, not neural networks.

Answer 2

Highly recommend the Stanford Introductory Machine Learning course on Coursera by Andrew Ng - a classic! Right from the start, in detail. In English.
As soon as you feel that he mentions something that is not entirely clear, google the topic and refresh your knowledge. Matrices, differentiation.

Answer 3

From books on deep learning with mathematics, I would recommend this one https://dmkpress.com/catalog/computer/data/978-5-9...

Answer 4

It is enough to parse the first part of the DL Book ( https://www.deeplearningbook.org/), looking at additional sources if necessary.