ИСТИНА

We study the problem of measuring photometric redshifts (photo-z) of extragalactic X-ray sources using machine learning techniques. The proposed photo-z models show accuracy (up to 2 times) better than the SOTA results for X-ray sources in the literature. The main accuracy improvement comes from using a large training sample (∼ 600k objects) and various features from several modern wide photometric surveys. The presented models are integrated into the SRGz system designed to construct a three-dimensional map of X-ray sources on the Eastern Galactic Hemisphere of the SRG/eRosita AllSky survey [1]. The SRGz system is developed in the science working group of RU eROSITA consortium on X-ray source detection, identification, and eROSITA source catalog in the High Energy Astrophysics Department at Space Research Institute of the Russian Academy of Sciences.