Is MatAn and Linear Algebra Enough for a Career in Machine Learning?

Question

Mathematics

The programming language is still outside the scope of the question, it is definitely on the agenda, and the question is about mathematics.

At the moment, I feel quite confident when it comes to the integral, (partial) derivatives, matrices, systems of linear algebraic equations, vectors and operations on them.
Also familiar with the basics of machine learning, in particular, these are the properties (features, parameters) of objects, weights, machine tasks. training (eg classification problem), some very simple algorithms.
I know some numerical methods: num. integration, differentiation, search for the extremum of a multidimensional function. And I know the basics of discrete mathematics and related theories of graphs, sets, Boolean algebra.
Now I'm learning probability theory and math. statistics.

However, delving into the topic of machine learning, I periodically stumble upon more complex and specialized topics that are visible in the formulas on presentations or in tutorials, but it is not entirely clear how they relate to this (machine learning). For example, such concepts as the "norm" of a vector or matrix, functional, Hesse and Jacobi matrices, positive definite matrices, and somewhere there was even a hint of a double integral.

All this is of course interesting, but it can still be spent a good couple of years on various kinds of n-dimensional topology, functional analysis and differential equations, but it is not known whether it will come in handy. It's not that I'm in a hurry, but I don't want to go through topics in vain that would be relevant mainly only to physicists and / or pure mathematicians.

It becomes not entirely clear whether my knowledge that I have listed is enough to work in the field of machine learning purely from a mathematical point of view, even if trying to read something on this topic leads to such problems? Maybe I'm just not reading the material for my level?
If something of the above is really worth knowing for me, recommend a book.

Answer 1

Lord, how many times can you ask the same thing. Poor qna.habr.
Well, here are the answers to your questions, study:
https://www.analyticsvidhya.com/blog/2019/10/mathe...
https://www.kdnuggets.com/2020/06/math-data-scienc. ..
https://towardsdatascience.com/the-roadmap-of-math...
https://proglib.io/p/obuchenie-data-science-kakie-...
https://habr.com/ru /post/432670/
https://www.datasciencecentral.com/profiles/blogs/...
https://qna.habr.com/q/1015602#answer_1980250

Answer 2

no special math is needed. A neural network is a system of fairly simple non-linear equations with a very large number of parameters (already billions!). This system can only be solved numerically by fitting parameters using multivariate optimization, a process called learning. To speed up, use the greedy learning method. The rest is art and mysticism