How to start learning machine learning?

Question

Node.js

Please point me in the right direction!
Input: self-taught, so there are gaps in knowledge, in general - php (before OOP, in particular before "factories") and js +/- at the same level. The topic of AI is very interesting, realizing that php is not a very suitable language for this, so I think to concentrate on learning node.js
What I want to understand at this stage:
- where and how to choose the right library to work for specific tasks
- how to work with these libraries
What I want to implement: in the Yandex promo video about CatBoost, weather forecasting is given as an example based on the analysis of previously collected data. I would like to implement a similar model based on numbers and the date they refer to. Those. there are only two types of transmitted information - number and date, but the sample is quite large.
What is the best place to start?
Which library should you pay attention to?
And in general, what can you advise?
Thank you!

Answer 1

How to start learning machine learning?

From opening Google and typing "machine learning, where to start". On the first page of the answers we find:
- links to half a dozen articles answering this question
- several links to lists of literature, video courses, online courses and just sites containing detailed annotations of everything. Take and choose what you like, interests, existing mathematical base.
- links to popular science portals dedicated to machine learning.
- and - here's the surprise - a link to Wikipedia, which has a wonderful "introductory course" on the topic and dozens - if not hundreds - of links to deepen understanding.
In general, there is so much information that it is time to ask the question - what NOT to use, and not where to find information on ML.

realizing that php is not quite a suitable language for this, so I think to concentrate on learning node.js

- well, this is also clearly not what is used in the field of AI and ML. But here the language is the last question to be solved. Because even PHP and JS already have libraries. True - not very powerful, but it is.

I would like to implement a similar model based on numbers and the date they refer to. Those. there are only two types of transmitted information - number and date, but the sample is quite large.

This task is from the section of statistics called "time series analysis". It has existed for almost a hundred years, and is solved by a dozen different methods. From linear regression to LSTM networks. And you need to start not with CatBoost or other newfangled tools, but with getting to know the base, gaining an understanding of how it all works and, most importantly, why. And only then what method to implement your forecast - decide effortlessly.

Which library should you pay attention to?

- Lord, let's not pay attention to libraries, but to algorithms and methods. And the library is only a technical tool for their implementation. Starting from the library is like a surgeon starting his training by choosing a scalpel model that he will operate on.
And finally - here are a couple of links for you, so as not to go far - directly to the topics of the same site, where a similar question was asked countless times. Read, think.
The data science path. How would be more correct?
Where to start learning artificial intelligence?
What topics in mathematics do you need to know to start studying machine learning?
What are the books on neural networks and AI?
The data science path. How would be more correct?