«Yandex» has opened to the public its system of machine learning

© RIA Novosti / Ramil to Setdiscoverable in photomontages of the company Yandex in Moscow«Yandex» has opened to the public its system of machine learning© RIA Novosti / Ramil to Setdiscoverable the image Bank

Programmers of «Yandex» has developed and put in open access system a machine learning CatBoost, which is already used in the Large hadron Collider, the press service of the company.

«Yandex many years of experience in machine learning, and CatBoost was created by the best specialists in this field. I must say that CatBoost is the first Russian machine learning technique, which became available to the public. I hope that the community will appreciate it and help make it better,» said Mikhail Bilenko, head of the Department of machine intelligence and research of «Yandex».

As the experts of «Yandex», the system machine learning is a direct descendant of the system Matriksnet, the basis of virtually all services of the Russian search giant. It is not a neural network but uses a special mathematical techniques, the so-called gradient boosting to assess the relevance of information and data analysis.

By this word the programmers understand the special set of algorithms, each of which gives the answer, not much different from random values, but their combination allows to obtain a sufficiently precise answer to the problem or to classify a particular object.

The main advantage of the algorithm CatBoost, according to programmers, is that it allows to use not only numeric data but also objects of another type, for example, pictures of clouds or types of buildings. This makes it easier for people using it to analyze and categorize information, and significantly expands its application.

In the experiment, it was used to improve the search results ranking of the tape of recommendations Yandex.Zen and the calculation of weather forecasts in technology Medium, and in all cases performed better Matrixnet.

Further CatBoost will work on other services. It also uses the Yandex Data Factory team in its decisions for the industry, in particular to optimize consumption of raw materials and predicting defects. In addition, CatBoost now uses CERN: he used the Russian algorithm to combine data obtained from different parts of the LHCb detector.

As explained by the press service of «Yandex», the company intends to use the same version of CatBoost in their work, which was made public, and the programmers plan to use community input to machine learning systems.


Be the first to comment on "«Yandex» has opened to the public its system of machine learning"

Leave a comment

Your email address will not be published.