What sections of mathematics are needed for machine learning?

Question

Mathematics

Good day!
I know what sections of mathematics are needed for ML. But what should you pay attention to first of all?
And is it worth learning everything? It needs a minimum. I don't remember mathematics since my studies.
I already got a job, with the condition that I learn.

Answer 1

Most likely, you will master the mashob with some kind of (video) course. The course description usually indicates what knowledge needs to be refreshed in order to study.
There are courses with high requirements for linear algebra, matan, mat. statistics. There are courses with a minimum of mathematics, where they try to explain it easier and show how to use it. The course will almost certainly be in English.
DeepLearning.ai , Udacity , Coursera , Yandex
So just choose the course that suits you best and go!
p.s. Congratulations on your work!

Answer 2

I want to clarify a little.
You write "I know what sections of mathematics are needed for ML." This is great, because most of the similar questions on this site come from people who didn't even bother to Google the answer themselves. Or at least on
https://qna.habr.com And writing an answer here for the hundred and twenty-eighth time is tiring. You have given yourself such work, ie. you carefully reviewed the list of at least a dozen answers to this question, analyzed them, understand what topics are open there and what they are for. This is great.
Here is incomprehensible only your question "is it worth learning everything?". Why, after reading these sources, you came to the conclusion that the people who wrote them did it in order to complicate your life and throw something into your answers that you do not need.
Now you want one of these answers to select "the minimum for this" for you. Those. - in fact, gave another answer to your question, which should reduce what others have already written well thought out?
But let's do it the other way around - you ask a question, and we answer you whether, for example, linear algebra or statistics is needed for you. It will be fair - you will show that you really understand what it is about, we will explain to you where ML is and what it is used for. And you yourself will decide, in your specific job, the one that you have already found, whether these tasks will occur or not. And they won’t kick you out because someone advised you that, for example, you don’t need knowledge of the laws of distribution. We don't know this.
And besides, keep in mind that each answerer looks at your question from the point of view of their own experience. To what extent is this view correct, i.e. how much you can believe the advice that will sound here is a very difficult question. If someone, for example, in their practice did not understand what optimization is, can we assume that this section is not needed?
PS Well, I was very embarrassed by this "I don't remember mathematics since my studies." - from a 3rd year student of the specialty "information systems".

Answer 3

First of all, you need a good command of the methods Mat. Statistics.
Further, probability theory or algebra, for example, but it depends on what tasks you are going to solve in ML.
Further, I would recommend mastering Math Statistics in practice in Python using the ML library sklearn for example.
Here are a bunch of examples - https://nbviewer.jupyter.org/github/Yorko/mlcourse...

Answer 4

You need, at a minimum, linear algebra and theoretical ver-statistics at a good level. Mathanalysis is enough school understanding. I would also highly recommend numerical methods - selectively. Let's say you use python libraries - you go into the documentation to see what algorithms are hardwired in them - and then you figure out how they work from the textbook. In standard courses, numerical methods are given little attention. (I teach mathematics in Gikbrains and I understand machine learning myself. I highly recommend my resource Kursopoisk, just my materials about mathematics for MI, for example, lin.alg www.kursopoisk.ru/?linear-algebra - everything is free and without registration.).