(Photo: NASA/Unsplash)
A global network of researchers has discovered 11 previously unknown cosmic anomalies by automating their searches using machine learning techniques. Seven of them are supernova candidates.
AI Discovers Cosmic Anomalies
Most astronomical discoveries have been derived from research and subsequent calculations. Although the overall number of observations in the 20th century remained relatively low, the amount of data increased significantly with the onset of large-scale astronomical research.
Manually analyzing such vast amounts of data is costly and time-consuming, so a team of SNAD researchers from Russia, France, and the United States worked together to create an automated solution.
The SNAD team is a consortium of researchers led by Matvey Kornilov, Associate Professor of Physics at HSE University. These researchers recently published their findings in the journal New Astronomy, outlining how to use machine learning to detect 11 previously unknown cosmic anomalies. Seven of these are believed to be supernova candidates.
Using a “nearest neighbor” approach, researchers used kD trees to discover anomalies in digital photographs of the northern sky collected in 2018. In computer science, a kd-tree (short for k-dimensional tree) is a spatial partitioning data structure. To organize points in k-dimensional space. With the help of AI, researchers were able to narrow down their search for cosmic anomalies.
Also read: Rapidly Expanding Black Hole Devouring Planet Discovered Using ANU’s SkyMapper Telescope
What is a space anomaly?
For the past few years, astronomers have studied celestial bodies by examining their light curves, showing how their brightness changes over time. They first perceive rays of light in the sky and track their progression to determine if the light brightens, dims, or dims over time.
According to Phys.org, the researchers evaluated one million real light curves from ZTF’s 2018 collection and seven simulated live curve models of the types of items reviewed in the study. did. So far, we’ve tracked a total of 40 factors, including the intensity and duration of object brightness.
According to co-author Konstantin Maranchev, they used a set of properties predicted to be detected in real objects to specify the quality of the simulation. For that purpose, I was looking for super-strong supernovae, type Ia supernovae, type II supernovae, and tidal disruption events in a collection of nearly one million items.
“We call such classes of objects anomalous. They are either very rare, have little-known properties, or appear interesting enough to warrant further study,” Malachev said. I got
We used the kD-tree algorithm to find the 15 nearest neighbors per simulation, the actual items from the ZTF database. A manual analysis validated 11 anomalies, of which 7 were possible supernovae and 4 were candidates for galactic nuclei in tidal disruption scenarios.
“In addition to the rare objects already discovered, we have detected several new objects that astronomers had previously missed,” said Maria Przynskaya, a co-author of the paper and a researcher at the Sternberg Institute for Astronomy. We were able to do this, which allows us to improve our algorithms so that existing searches don’t miss such objects.”
This study reveals that SNAD’s kD-tree and AI-machine algorithm strategies are successful and fairly easy to implement. The techniques proposed to identify specific forms of cosmic events are universal and may be used to find fascinating astronomical phenomena, not just rare kinds of supernovae.
Related article: Orbital problems experienced by Momentus: Space Tugs error
This article is owned by Tech Times
Written by Thea Felicity
ⓒ 2022 TECHTIMES.com All rights reserved. Please do not reproduce without permission.
.
Comments
Post a Comment