Hyderabad: Artificial Intelligence (AI) and Machine Learning (ML) are continuing to help researchers work, sorting large quantities of data to help them reach results faster than ever. These technologies are now helping astronomers find supernovas in sky full of space noise.
The supernova is a rare and bright explosion indicating the death of a giant star. However, identifying them is not so easy as they must be discovered immediately before fading, while distinguishing them from other device errors and known objects. A team of researchers, led by Oxford and Queens University, will search for these using the Asteroid Terrestrial Impact Trastor Alt System (ATLAS), a project funded by NASA and led by the University of Hawaii.
Originally built as an asteroid impact early warning system, Atlas uses five telescopes around the world to scan the entire visible sky every 24-48 hours. Oxford processes data for high explosions across our galaxy, bringing millions of potential alerts each night. The researchers then apply standard filtering and automated image analysis techniques, leaving 200-400 candidate signals that need to be manually checked and take several hours each day.
To tackle this issue, Oxford created an AI tool called Virtual Research Assistant (VRA). This can probably use the ML approach to identify several true signals caused by supernovae, via thousands of data alerts. According to the survey, Astrophysical Journalreduces astronomer workload by up to 85%.
The new tool is a collection of automated bots that mimic human decision-making processes by ranking alerts based on realistic, galaxy explosion possibilities. Unlike data-hungry deep learning models that require vast amounts of training data and supercomputers, VRA uses a more lean approach, using small algorithms based on decision trees, looking for patterns in selected aspects of the data. This allows scientists to use their expertise in training and steer them to look for the key features they need for their tasks.
“The surprising thing is how little data is. With just 15,000 examples and the computing power of a laptop, we can train smart algorithms to do heavy lifting and automate what happens every day humans have hours.”
VRA also updates its ratings every time the telescope revisits the same sky patch. This means that the signal is automatically checked and rescores over several nights. This will lead human astronomers to review only the most promising candidates.
In the first year of use, VRA successfully filtered over 30,000 alerts, but less than 0.08% of actual supernova alerts were missing, retaining more than 99.9% of real supernova candidates. In December 2024, researchers linked the tool to a South African Lesedi telescope, alerting automated “promising signal” alerts, even before humans review the data.
Researchers say that VRA will be examining the entire Southern Hemisphere every few days for over a decade, as the Vera Rubin Observatory's Legacy Survey of Space and Time (LSST) is preparing for its launch in early 2026, and will eventually generate over 500 petabytes (1 PB = 1,024 TB).
