Machine learning is a pervasive term, but not everyone gets it right. In this article, the experts of the direction of analytical solutions of KORUS Consulting Group Alena Gaybatova and Ekaterina Stepanova will tell you what machine learning (ML) really is, in which cases this technology should be used in projects, and also where machine learning is actively used in practice ...
How data is handled
For a long time, at meetings with customers, we began to notice that everyone confuses machine learning, artificial intelligence (AI), big data and other terms from this field.
So, the general name of the technology is artificial intelligence. It is of two types - strong (aka general) and weak. We won't be discussing strong AI especially as these are Terminator-level decisions. We are slowly approaching it, but so far it exists only in the form of fragments of weak AI collected together (as, for example, in "smart" columns).
It's much more interesting to talk about weak artificial intelligence. It is also divided into two types. The first is expert systems , manually programmed algorithms (for example, a linguist's group programmed an algorithm for translating words from one language into another).
The second is the so-called data-driven systems that extract the logic of work from some kind of historical data. This type has many synonymous terms that have arisen over time:
fashionable in the 90s and zero data mining and knowledge discovery from database (KDD),
data science, which came into use closer to 2010,
big data is popular today. The only exception, or rather an addition, which this very term introduces is the presence of a huge amount of complexly structured data.
Different algorithms for different tasks
In accordance with the two types of weak AI, we can draw conclusions from the data manually (with expert systems) and using machine learning. It, in turn, is subdivided into two types: classic ML and deep learning (using deep neural networks with a large number of layers).
ML . , , : , . , , c.
— , , . , : , , . , . , .
, — . , , .
? : , . : — 3 . , .
- , , .
– , , . — : , , — , .
— . , - . — , , , . . . , .
, .
– . C . 4 . – . , , . , , . .
C , — .
1950- . , : .
, . , . , . , , , .
ML . .
, . , .
. ( – , , ). , , , , , .
Uplift, . Uplift , , ( , ), .
–
, . ML . , , .
, - ;
, / ;
, , .
– . C ,
, , , , , . - .
, ML , , , , , , , , — .
, . , , , , , , . , , .
— . , , , , . – , , , . ML — , , , , , , .
, , . ( ), . , , – , - .
C , , , , , , , .
, , , . , , , ML- -.
, , . , , , . , .