Artificial created in Portugal is ready for intelligence to help identify astronomical objects – Multimédia
Catalog astronomical objects is a task that must be challenging, due to the amount of objects and the complexity of the Universe, but artificial intelligence is proving to be a very promising tool for solving this problem.
This is how researchers from the Institute of Astrophysics and Space Sciences Pedro Cunha and Andrew Humphrey who created the SHEEP, a world of artificial intelligence (machine learning) that identifies the nature of.
SHEEP is a supervised artificial intelligence pipeline that estimates photometric redshifts and later uses this information to classify astronomical objects as astronomical, quasars, or stars.
“Photometric information is the easiest to obtain and that is why it is very important in an analysis of the nature of the objects observed”, says Pedro Cunha, a PhD student at the IA and at the Department of Physics and Astronomy at the Faculty of Sciences of the University of Porto , the first author of the article now published in the journal Astonomy & Astrophysics.
One of the novelties of the process is that, before starting the classification, SHEEP first estimates photometric redshifts, information that is added to the data, as an additional feature for learning the classification modelexplains the Institute of Astrophysics and Space Sciences in a statement.
A device discovered that, by including this data along with the coordinates (right ascension and declination) of objects, an artificial intelligence was able to 3D Universe and this knowledge, parallel with color information, perceive to make more accurate estimates of object properties. For example, the reserve feature that there is a higher probability of finding stars closer to the Milky plane than at the galactic poles.
“When we gave artificial intelligence a three-dimensional view of the Universe, we greatly increased its ability to more accurately decide what kind of celestial object it was cataloging.,” adds Andrew Humphrey.
Several surveys of the sky, such as the Sloan Digital Sky Survey (SDSS), producing large amounts of data that revolutionize an area of astronomy and future surveys, which will be carried out by instruments such as the Vera C. Rubin Observatory, the Dark Energy Spectroscopic Instrument (DESI), each Euclid space mission or the James Web space turn continue to propose the most space space images.
Click on the images for more details
Due to its huge amount, it takes a long time to get this data through traditional methods. New artificial intelligence methods are considered to analyze and take better advantage of this new data.
This work integrates the equipment to take advantage of the wealth of information that these tests will produce, when developing machine learning systems that can classify and characterize billions of sources, efficiently.
“One of the most exciting things is seeing how this animal is helping us to better understand the Universe.”, says Pedro Cunha. “Our method showed us a possible path, with new ones being created during the process. It’s an exciting era for astronomy”.
Photometric and spectroscopic research are one of the main ways of understanding the visible content of the Universe. The data from these objects allowed statistical studies of stars, quasars and astronomy, in addition to the discovery of more peculiar astronomical objects.