Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

The miniJPAS and J-NEP surveys: Machine learning for star-galaxy separation

Ana Paula Jeakel, Gabriel Vieira dos Santos, Valerio Marra, Rodrigo von Marttens, Siddhartha Gurung-López, Raul Abramo, Jailson Alcaniz, Narciso Benitez, Silvia Bonoli, Javier Cenarro, David Cristóbal-Hornillos, Simone Daflon, Renato Dupke, Alessandro Ederoclite, Rosa M. González Delgado, Antonio Hernán-Caballero, Carlos Hernández-Monteagudo, Jifeng Liu, Carlos López-Sanjuan, Antonio Marín-Franch, Claudia Mendes de Oliveira, Mariano Moles, Fernando Roig, Laerte Sodré, Keith Taylor, Jesús Varela, Héctor Vázquez Ramió, José M. Vilchez, Christopher Willmer, Javier Zaragoza-Cardiel

TL;DR

This paper tackles robust star-galaxy separation in the J-PAS Pathfinder data (miniJPAS and J-NEP) by building a large labeled dataset via multi-survey crossmatches and training supervised ML models. An XGBoost classifier is optimized with TPOT and evaluated on photometric and morphological features; results show morphology improves performance over photometry alone. Photometry-only yields $AUC_{ROC} \approx 0.992$, while morphology+photometry yields the best $AUC_{ROC}$, and permutation analysis identifies $c_r$, $\mu$, and PSF as key morphological differentiators and bands around $3900$–$6800$ Å as informative. The study provides a value-added catalog (VAC) with the trained models, enabling robust downstream analyses in cosmology and galaxy evolution with reliable star-galaxy classifications.

Abstract

We present a supervised machine learning classification of sources from the Javalambre Physics of the Accelerating Universe Astrophysical Survey (J-PAS) Pathfinder datasets: miniJPAS and J-NEP. Leveraging crossmatches with spectroscopic and photometric catalogs, we construct a robust labeled dataset comprising 14594 sources classified into extended (galaxies) and point-like (stars and quasars) objects. We assess dataset representativeness using UMAP analysis, confirming broad and consistent coverage of feature space. An XGBoost classifier, with hyperparameters tuned using automated optimization, is trained using purely photometric data (60-band J-PAS magnitudes) and combined photometric and morphological features, with performance thoroughly evaluated via ROC and purity-completeness metrics. Incorporating morphology significantly improves classification, outperforming the baseline classifications available in the catalogs. Permutation importance analysis reveals morphological parameters, particularly concentration, normalized peak surface brightness, and PSF, alongside photometric features around 4000 and 6900 A, as crucial for accurate classifications. We release a value-added catalog with our models for star-galaxy classification, enhancing the utility of miniJPAS and J-NEP for subsequent cosmological and astrophysical analyses.

The miniJPAS and J-NEP surveys: Machine learning for star-galaxy separation

TL;DR

This paper tackles robust star-galaxy separation in the J-PAS Pathfinder data (miniJPAS and J-NEP) by building a large labeled dataset via multi-survey crossmatches and training supervised ML models. An XGBoost classifier is optimized with TPOT and evaluated on photometric and morphological features; results show morphology improves performance over photometry alone. Photometry-only yields , while morphology+photometry yields the best , and permutation analysis identifies , , and PSF as key morphological differentiators and bands around – Å as informative. The study provides a value-added catalog (VAC) with the trained models, enabling robust downstream analyses in cosmology and galaxy evolution with reliable star-galaxy classifications.

Abstract

We present a supervised machine learning classification of sources from the Javalambre Physics of the Accelerating Universe Astrophysical Survey (J-PAS) Pathfinder datasets: miniJPAS and J-NEP. Leveraging crossmatches with spectroscopic and photometric catalogs, we construct a robust labeled dataset comprising 14594 sources classified into extended (galaxies) and point-like (stars and quasars) objects. We assess dataset representativeness using UMAP analysis, confirming broad and consistent coverage of feature space. An XGBoost classifier, with hyperparameters tuned using automated optimization, is trained using purely photometric data (60-band J-PAS magnitudes) and combined photometric and morphological features, with performance thoroughly evaluated via ROC and purity-completeness metrics. Incorporating morphology significantly improves classification, outperforming the baseline classifications available in the catalogs. Permutation importance analysis reveals morphological parameters, particularly concentration, normalized peak surface brightness, and PSF, alongside photometric features around 4000 and 6900 A, as crucial for accurate classifications. We release a value-added catalog with our models for star-galaxy classification, enhancing the utility of miniJPAS and J-NEP for subsequent cosmological and astrophysical analyses.

Paper Structure

This paper contains 33 sections, 3 equations, 20 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Data
  3. Crossmatch with SDSS DR12
  4. Crossmatch with Gaia EDR3
  5. Crossmatch between miniJPAS and DEEP3 DR4
  6. Crossmatch between J-NEP and Binospec data
  7. Crossmatch between miniJPAS and DESI DR1
  8. Crossmatch between miniJPAS and HSC-SSP PDR2
  9. Final labeled set
  10. Feature configuration
  11. Observational features
  12. Photometric features
  13. Morphological features
  14. Feature sets
  15. Supervised classification
  16. ...and 18 more sections

Figures (20)

  • Figure S1: Crossmatch between DEEP3 DR4 and miniJPAS sources with $r < 23.5$.
  • Figure S2: Crossmatch between Binospec and J-NEP sources with $r < 23.5$.
  • Figure S3: Crossmatch between DESI DR1 and miniJPAS sources with $r < 23.5$.
  • Figure S4: $r$-band magnitude distributions of labeled stars and galaxies across the catalogs used for crossmatching. SDSS spectroscopic and photometric classifications are separated at $r=20$.
  • Figure S5: $r$-band magnitude distributions of labeled stars and galaxies in miniJPAS and J-NEP.
  • ...and 15 more figures