(ORDO NEWS) — The classification of celestial objects is a longstanding astronomical problem. It is easy to confuse sources of different types located at unimaginably great distances, for example, a star with a galaxy, or a quasar with a supernova.
One of the modern solutions to this problem was recently proposed by researchers led by Pedro Cunha from the Institute of Astrophysics and Space Research, Portugal, who tried to solve the classic problem by creating an original machine learning algorithm called SHEEP, capable of determining the nature of astronomical sources.
The SHEEP algorithm is an AI-based data-processing program that estimates photometric redshifts and uses this information to further classify sources such as a galaxy, quasar, or star. “This photometric information is easy to obtain and therefore very important in the initial analysis of the nature of observed objects,” Cunha said.
“The novelty of our study is that before performing the classification, the algorithm first estimates the photometric redshifts, which are then added to the data set as an additional auxiliary parameter for training the classification model.”
The team found that adding this redshift and the coordinates of the objects allowed the AI to “understand” these objects in the grid of the 3D map of the universe, and they used these values in conjunction with color information to make more accurate estimates of the properties of the sources.
For example, artificial intelligence has found that the probability of finding stars closer to the plane of the Milky Way is higher than in the direction of the poles of the Galaxy.
According to the authors, when artificial intelligence began to work with a three-dimensional map of the Universe, the accuracy of determining the nature of sources increased significantly.
This work has become an important part of modern attempts to develop means of processing the huge amount of data produced by today’s operational sky surveys, such as the Sloan Digital Sky Survey, as well as planned projects such as the Vera Rubin Observatory, Dark Energy Spectroscopic Instrument (DESI), the Euclid satellite ( Euclid and NASA/ESA‘s James Webb Space Observatory.
—
Online:
Contact us: [email protected]
Our Standards, Terms of Use: Standard Terms And Conditions.