COSMOS2025: Machine Learning Classification of Early- and Late-type Galaxies at 0 < z < 3 0<\text{z}<3 Matplotlib ( 9 ) , Pandas ( 61 ) , Scikit-learn ( 68 ) , Astropy ( 75 ) , UMAP ( 60 ) , Numpy ( 36 ) , SciPy ( 91 ) , Seaborn ( 93 ) .
Vahid Asadi, Najmeh Sheikhi · 2026
Read COSMOS2025: Machine Learning Classification of Early- and Late-type Galaxies at 0 < z < 3 0<\text{z}<3 Matplotlib ( 9 ) , Pandas ( 61 ) , Scikit-learn ( 68 ) , Astropy ( 75 ) , UMAP ( 60 ) , Numpy ( 36 ) , SciPy ( 91 ) , Seaborn ( 93 ) . by Vahid Asadi, Najmeh Sheikhi (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-03224.txt.
Description
Asadi and Sheikhi sort galaxies into early and late types with a fast machine learning method. The data come from the COSMOS2025 catalog, at redshifts from zero to three. Past routes need image labels or slow structural fits. This team trains on simulated galaxies instead, with labels from a semi-analytic model.
Labels rest on bulge-to-total mass ratio and star formation rate. Real-style photometric noise is added to bridge the gap to true data. A CatBoost model trained on 66 broadband colors did well in the simulated domain. It caught late types at about 98 percent precision and recall. Early types scored lower but still well. Applied to 44,132 real galaxies, the model put about 94 percent of cases at high confidence. Checks against bulge-disk fits gave 70 percent overall accuracy. The whole pipeline runs in under 30 minutes on standard hardware.
Catalog record
| CREATOR | Vahid Asadi, Najmeh Sheikhi |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 60 KB |
| LICENSE | http://creativecommons.org/publicdomain/zero/1.0/ |
| RIGHTS | CC0 |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2606.03224