Aussie scientists develop AI program to spot far-off galaxies

Source: Xinhua| 2018-10-31 13:43:57|Editor: Yurou
Video PlayerClose

SYDNEY, Oct. 31 (Xinhua) -- Australian scientists on Wednesday said they have trained an artificial intelligence (AI) program used to recognize faces on social networking site Facebook to also spot galaxies in deep space and better probe the universe.

The AI program scans images taken by radio telescopes and its job is to spot galaxies that emit powerful radio jets from "supermassive black holes" at their centers, said the University of Western Australia researchers in a statement explaining their project.

The jets "can stretch a long way from their host galaxies" over time and make it difficult for traditional computer programs to figure out where the galaxy is - gaps which the AI bot aims to plug, said the university's astronomer Dr Ivy Wong.

The program grew out of an open source version of tech giant Microsoft and Facebook's object detection software and was completely overhauled and trained to recognize galaxies instead of people, said the university's big data specialist Dr Chen Wu.

The AI taps highly accurate databases and an extensive neural network to help spot the radio jets, in line with a new paradigm of programming, said the researchers.

"The new generation of programmers spend 99 percent of their time crafting the best quality data sets and then train the AI algorithms to optimize the rest," said Wu.

"This is the future of programming."

Traditional computer algorithms are able to correctly identify 90 percent of the sources and that "still leaves 10 percent, or seven million 'difficult' galaxies that have to be eyeballed by a human due to the complexity of their extended structures," said Wong.

The latest AI technology "has huge implications for how telescope observations are processed," she said. A research paper on the program was published in scientific journal Monthly Notices of the Royal Astronomical Society.

"If we can start implementing these more advanced methods for our next generation surveys, we can maximize the science from them," said Wong.