Discovering Hidden Galaxy Mergers: Finding Mergers Missed by Detection Methods

Overview

When two galaxies collide, they should leave behind visible scars. But some collisions happen in massive, old galaxies where the evidence fades so fast that every detection tool declares them normal. I built a machine learning algorithm that looks at whether the outer edges of a galaxy are more disturbed than the center, a signal that standard tools never measure. Applied to 66,018 real galaxies from 20 years of telescope data, it found 301 hidden collisions that every existing tool had missed, and a separate project where thousands of human volunteers classified the same galaxies independently confirmed the discoveries. Every study since 2000 that counted galaxy collisions has been systematically missing the most massive ones, biasing our models of how the universe built its largest structures.

Awards (2)

• Bronze Medal
• Selected for CWSF 2026

Competition history

• CWSF 2026 Aerospace

Related projects

Closest projects by meaning, across every fair and year in the corpus.