Azernews.Az

By Alimat Aliyeva

Astronomers using artificial intelligence have discovered more than 1,300 rare astrophysical objects hidden in archived data from the Hubble Space Telescope, Azernews reports.

The team analyzed nearly 100 million small images from the Hubble Legacy Archive in just two and a half days. More than 800 of the detected objects had never been previously described in scientific literature, highlighting how much valuable information can remain unnoticed in existing datasets.

Among the unusual findings are merging and interacting galaxies, gravitational lenses, galaxies hosting massive bursts of star formation, and so-called “jellyfish galaxies” with long gas “tentacles” stripped by their environment. The researchers also identified protoplanetary disks whose shapes resemble hamburgers, as well as several dozen objects that do not fit into any current astronomical classification.

To tackle this challenge, researchers at the European Space Agency (ESA), David O’Ryan and Pablo Gómez, developed a neural network called AnomalyMatch. The system is designed to recognize rare and unexpected features in astronomical images, mimicking how the human brain processes visual information but doing so at vastly greater speed.

Manually examining such an enormous volume of data would be virtually impossible. After the algorithm flagged the most unusual candidates, the scientists carefully reviewed the results and confirmed more than 1,300 genuine anomalies.

Tools like AnomalyMatch are expected to become essential for future astronomical research. Upcoming observatories — including the Nancy Grace Roman Space Telescope, the Euclid mission, and the Vera C. Rubin Observatory — will produce unprecedented amounts of data, far beyond what humans alone can analyze.

Some of the “anomalies” identified by AI today could represent entirely new classes of cosmic objects, suggesting that the universe may still be hiding phenomena we haven’t even imagined yet.