Particle.news
Download on the App Store

ESA AI Mines Hubble Archive, Uncovers 1,300 Cosmic Anomalies—800 Newly Identified

The AnomalyMatch system completed the archive’s first systematic anomaly hunt in days and now positions astronomers to tackle far larger surveys from Euclid, Rubin and NASA’s Roman telescope.

Overview

  • ESA researchers David O’Ryan and Pablo Gómez built the neural network AnomalyMatch to detect rare objects in decades of Hubble data.
  • Scanning nearly 100 million image cutouts in about two and a half days, the tool produced a ranked list that human review confirmed as more than 1,300 true anomalies.
  • Finds include numerous galaxy mergers and interactions, gravitational lenses, jellyfish and ring or clumpy galaxies, edge-on planet-forming disks, and several dozen objects that resist classification.
  • This work constitutes the first systematic anomaly search of the Hubble Legacy Archive and is documented in a peer-reviewed Astronomy & Astrophysics paper.
  • The team highlights the method’s scalability for upcoming data-rich missions, with detailed classification and scientific follow-up of the most intriguing candidates to come.