ESA astronomers using AI discovered over 800 new cosmic anomalies in Hubble images
Astronomers from the European Space Agency (ESA) discovered over 800 previously unknown cosmic anomalies using a new artificial intelligence-based tool. The neural network named AnomalyMatch analyzed almost 100 million image fragments from the Hubble telescope archive in less than three days.
What is known
The algorithm was developed by researchers David O'Ryan and Pablo Gomez. It was trained to work with the Hubble Legacy Archive — a database containing tens of thousands of observation sets from the telescope's 35 years of operation. ESA notes that the volume of data is so large that manual analysis with the required detail is practically impossible.
After the initial selection, the results were verified by scientists. Overall, the system identified about 1400 anomalous objects, of which more than 800 were not documented before. Most of the findings are galaxies that are interacting or merging, as well as gravitational lenses, where the light of distant objects is distorted by massive foreground galaxies.
Additionally, researchers recorded protoplanetary disks, "jellyfish-like" galaxies, and objects with clusters of stars of unusual shape. Several dozen findings could not be attributed to known classes at all. According to the project authors, AnomalyMatch could become an effective tool for analyzing large astronomical archives in the future.
Source: NASA
