Categories: world

AI trained to recognize deep galaxies

The researchers restored a Facebook-AI recognizing people on photos to identify deeper galaxies instead. The new boat, named ClaRAN, scans…

The researchers restored a Facebook-AI recognizing people on photos to identify deeper galaxies instead.

The new boat, named ClaRAN, scans radio telescope images in the hope of detecting radiolaxes, delivering powerful radio rays from a super massive black hole (SMBH).

ClaRAN is the brainchild of great computer specialist Chen Wu and astronomer Ivy Wong, both from the University of Western Australia, nod from the International Center for Radio Astronomy Research (ICRAR).

Super massive black holes are at the center of almost all known massive galaxies. As for our Winter Street, SMBH corresponds to the location of Sagittarius A *.

According to Wong, these black holes sometimes put burp out rays that can be seen with a radio telescope.

“Over time, the rays can extend far from their host galaxies, making it difficult for traditional computer programs to figure out where the galaxy is,” she explained. “That’s what we’re trying to teach ClaRAN do.”

Based on an open source version of Microsoft and Facebook’s object detection program, the reworked program is trained to recognize galaxies rather than people.

ClaRAN is also open source and widely available at GitHub.

By combining data from different telescopes, ClaRAN’s “confidence level” increases in its detections and classifications. Displayed as the figure above the detection box, shows a confidence of 1.00. ClaRAN is very convinced that the source detected is a radio-arc beam system and it has been properly classified (via Chen Wu & Ivy Wong / ICRAR / UWA)

We are currently aware of about 2.5 million radio sources, but expect to reveal another 70 million with the forthcoming Evolutionary Map of the Universe (EMU) survey.

EMU is a major project that will use the new Australian square-kilometer Array Pathfinder (ASKAP) telescope to make a census of radio sources in the sky. Traditional computer algorithms should be able to correctly identify 90 percent of these sources.

“Still, 10 percent or 7 million” difficult “galaxies remain to be eclipsed by a human because of the complexity of their expanded structures.” Wong said.

“If ClaRAN reduces the number of sources that require visual classification down to 1 percent, it means more time for our citizen researchers to spend watching new types of galaxies,” she added.

Wong previously used human power to detect galaxies through the popular Radio Galaxy Zoo project.

Volunteers from the group helped to produce the catalog used to educate ClaRAN-an example of a new paradigm called Wu “Programming 2.0”.

“Everything you make a large network, give it lots of data and let it find out how to customize its internal connections to generate the expected result,” he says. “This is the future of programming.”

One Research papers about ClaRAN were released today in the magazine Monthly Announcements from the Royal Astronomical Society published by Oxford University Press.

More Cosmic Coverage on :

Published by