This article is more than 1 year old
AI beats astroboffins at sniffing out fast radio bursts amid the universe's clutter
Who knows if the signals are signs of extaterrestrial activity?
AI is helping astronomers spot fast radio bursts, a mysterious class of signal emitted from a new type of object very rarely found in space that boffins are still trying to classify.
Fast radio bursts (FRBs) are difficult to study. They don’t crop up too often - there have only been around 30 confirmed events since their discovery over a decade ago - and the energetic flares are merely blips in the open sky, lasting only a few milliseconds.
Luckily, scientists have managed to hone in one of the most intriguing sources found yet. Codenamed FRB 121102, it’s three billion light years away from Earth and is the only site that is currently known to repeatedly send out these radio bursts.
A group of researchers decided to knuckle down and study FRB 121102 as part of the Breakthrough Listen project set up by the SETI Institute, a non-profit research organization focused on finding extraterrestrial intelligence
They analysed data taken from Green Bank Telescope observatory in Virginia to see if they could find any more bursts that they might have missed using a convolutional neural network (CNN). The system was fed a series of spectrograms that show the frequency and duration of a signal. Since real FRBs are rare, the researchers simulated data by creating a training dataset of around 400,000 images, half contain simulated pulses and the half do not.
These fake spectrograms were then fed into the CNN so it could learn what patterns and characteristics are most likely to be from FRBs. When the CNN was given real data from the Breakthrough Listen project, it found 72 new signals from FRB 121102.
“Together with the 21 previously reported pulses, this observation marks the highest number of FRB 121102 pulses from a single observation, totaling 93 pulses in five hours, including 45 pulses within the first 30 minutes,” according to the paper accepted in the Astrophysical Journal (here is a free version).
“This work is exciting not just because it helps us understand the dynamic behavior of fast radio bursts in more detail, but also because of the promise it shows for using machine learning to detect signals missed by classical algorithms,” said Andrew Siemion, co-author of the paper and a director of the Berkeley SETI Research Center and principal investigator for Breakthrough Listen.
AI and astronomy is a thriving area of research. Neural networks have been used to check if galaxies are still actively forming stars, measure the probability of extraterrestrial life on exoplanets, find gravitational lenses, and even super rare hypervelocity stars.
The new data will help scientists better understand the origin of FRBs. There are many ideas as to what the object could be. Some believe it could be a new type of supernova, neutron stars or even signs of alien life.
“Whether or not FRBs themselves eventually turn out to be signatures of extraterrestrial technology, Breakthrough Listen is helping to push the frontiers of a new and rapidly growing area of our understanding of the Universe around us,” said Gerry Zhang, co-author of the paper and a PhD. student at the University of California, Berkeley. ®
Biting the hand that feeds IT
