The researchers applied deep learning techniques to a previously studied dataset of nearby stars and revealed eight previously unidentified signals of interest. The team of researchers was led by University of Toronto undergraduate Peter Ma, along with those from the SETI Institute, Breakthrough Listen and scientific research institutions around the world, the study said. The research is published in the journal Nature Astronomy.
When considering the likelihood of discovering technologically advanced extraterrestrial life, the question often arises: “if they’re out there, why haven’t we found them yet?” And often the answer is that we’ve only searched a small part of the galaxy. Further, algorithms developed decades ago for the earliest digital computers may be outdated and inefficient when applied to modern petabyte-scale data sets, the study said.
Peter Ma, lead author says “In total, we searched 150 TB of data on 820 nearby stars on a dataset previously searched in 2017 with classical techniques but flagged as lacking interesting signals”.
Are we alone?
With the MeerKAT telescope today, we are extending that search to 1 million stars and beyond. We believe that work like this will help accelerate the speed at which we are able to make discoveries in our great effort to answer the question ‘are we alone’ in the universe?
The Search for Extraterrestrial Intelligence (SETI) looks for evidence of extraterrestrial intelligence originating on Earth by trying to detect technosignatures or evidence of technology that extraterrestrial civilizations may have developed.
Deep learning techniques to a classic search algorithm
The most common technique is to search for radio signals. Radio is a great way to send information across incredible interstellar distances; it quickly passes through the dust and gas that permeates space, doing so at the speed of light, about 20,000 times faster than our best rockets, Many SETI efforts use antennas to listen for any radio signals the aliens may be transmitting, the study said.
According to the study, it re-examined data taken by the Green Bank Telescope in West Virginia as part of the Breakthrough Listen campaign, which initially did not list any targets of interest. The goal was to apply new deep learning techniques to a classic search algorithm to achieve faster and more accurate results.
After running the new algorithm and manually re-examining the data to confirm the results, the study found that the newly detected signals had several key characteristics:
1) The signals were narrowband, meaning they had a narrow spectral width of only a few Hz. Signals caused by natural phenomena tend to be broadband.
2) The signals had a non-zero drift rate, meaning the signals had a slope. Such slopes could indicate that the origin of the signal had some relative acceleration with our receivers and is therefore not local to the radio observatory.
3) Signals appeared in source-on observations and not in source-off observations. If the signal comes from a certain celestial source, it will appear when we point the telescope towards the target and disappear when we look away. Human radio interference usually occurs in ON and OFF observations because the source is close.
These results dramatically illustrate the power of applying modern machine learning and computer vision methods to data challenges in astronomy, resulting in both new detections and increased performance.
Cherry Ng, another of Mao’s research advisers and an astronomer at both the SETI Institute and the French National Center for Scientific Research says “The large-scale application of these techniques will be transformative for radio signature science”.
While re-examination of these new targets of interest has yet to lead to the re-detection of these signals, this new approach to data analysis may allow researchers to more efficiently understand the data they collect and act quickly to re-examine the targets.
Written by: Vaishali Verma
