Artificial Intelligence used to discover gravitational lens events

Bringing artificial intelligence into astronomy will quicken the painstaking process of searching through piles of observations


Gravitational lensing is the result of light being bent by an extremely massive object. Image credit: NASA/ESA

Astronomers from a combination of universities including Groningen, Naples and Bonn have developed an exciting new way to sift through large amounts of telescope observations using Artificial Intelligence (AI). This form of AI is an algorithm that is similar to the one used by Google, Facebook and Tesla. Thanks to this AI algorithm, the team of astronomers have now discovered 56 new gravitational lens candidates.

A gravitational lensing event occurs when a galaxy is hidden behind another galaxy, but because the intermediate galaxy’s mass is so massive it bends the light from its remote galactic partner. This is in accordance to Einstein’s general theory of relativity, and these events provide astronomers with vital informational regarding the nature of dark matter. This is one of the aspects the Kilo-Degree Survey has been studying, but the only problem is that the survey produces images faster then we can process them. Even with the help of the general public, the astronomers can’t keep up with the pace.

56 new gravitational lens candidates have now been discovered, thanks to the help of artificial intelligence. Image credit: C. E. Petrillo (University of Groningen)

This is when the astronomers decided to introduce, what is called, ‘convolutional neural networks’. To put this in layman’s terms, it is a form of machine learning that analyses visual images. Google has used neural networks to win a match of Go against the world champion, Facebook uses this network to recognise what is in images on timelines and Tesla use it in their pursuit of self-driving cars.

“This is the first time a convolutional neural network has been used to find peculiar objects in an astronomical survey,” says Carlo Enrico Petrillo of the University of Groningen. “I think it will become the norm since future astronomical surveys will produce an enormous quantity of data which will be necessary to inspect. We don’t have enough astronomers to cope with this.”

In this instance, the neural network went through millions of images from the Kilo-Degree Survey that only covered 255 square degrees of sky; this is equivalent to only half a percent of the sky. The results found 761 gravitational lens candidates, 56 of which were confirmed to be candidates by the astronomers. Whether these are actual gravitational lenses will need further confirmation from telescopes such as the Hubble Space Telescope. During this search, the neural network did rediscover two known gravitational lenses, but it missed one because of its small size. Astronomers and researchers will continue to improve this network.

Keep up to date with the latest reviews in All About Space – available every month for just £4.99. Alternatively you  can subscribe here for a fraction of the price!

Tags: , , , , , , ,