A machine learning program, ASTRONOMALY, has scanned four million galaxy images, discovering 1635 anomalies, including 18 previously unidentified ones. The program, which operates unsupervised, can identify new types of outliers, such as gravitational lenses and galactic mergers. However, it performs best when humans correct its mistakes, a process known as active learning. This combination of AI and human input promises to revolutionize astronomy, enabling the analysis of vast amounts of data from large wide-field surveys.
Read more at Universe Today…
