MergeFinder: A Deep Learning Model for Galaxy Merger Identification
ISEF · 2024 Physics and Astronomy
Overview
Galaxy mergers play an integral role in galaxy formation and evolution, and with many large upcoming surveys, a reliable and efficient method of identifying mergers is crucial to their study. Currently, mergers are manually classified through visual inspection, which is incredibly time-consuming and biased towards major mergers (those with mass ratios less than 3:1) due to their conspicuous nature. This study developed MergeFinder, a convolutional neural network for galaxy merger identification. MergeFinder was trained on simulation data from the Evolution and Assembly of GaLaxies and their Environments (EAGLE) database, which contains physics-informed models of various types of mergers from different orientations. This enables MergeFinder to identify mergers based on subtle structural asymmetries, as opposed to solely the presence of visible companions. MergeFinder also utilizes a novel sequential classification technique, where it first classifies galaxies by orientation and merging status, then combines the orientation classes into one merging class and one non-merging class; this allows for higher performance than traditional binary classifiers. MergeFinder achieved an accuracy of 94%, a sensitivity of 93%, and a specificity of 96%. In addition, this research used MergeFinder to suggest possible reclassifications for existing galaxy data from the Dark Energy Camera Legacy Survey (DECaLS) and found that at least 12.8% of its current classifications are incorrect.
Competition history
- ISEF 2024
Resources
Related projects
CWSF · 2026
Discovering Hidden Galaxy Mergers: Finding Mergers Missed by Detection Methods
ISEF · 2017
A Linked Learning Approach to Automated Galaxy Morphology Classification
ISEF · 2022
Analysis of Ring Galaxies Detected Using Deep Learning With Real and Simulated Data
ISEF · 2025
Deep Learning Methods for Galaxy Morphology Classification
Closest projects by meaning, across every fair and year in the corpus.
Source: Regeneron International Science and Engineering Fair