J-PAS: Semi-Supervised Sim-to-Obs Transfer for Robust Star--Galaxy--Quasar Classification
Abstract
Modern studies in astrophysics and cosmology increasingly rely on simulations and cross-survey analyses, yet differences in data generation, instrumentation, calibration, and unmodeled physics introduce distribution mismatches between datasets (``domain shift''). In machine-learning pipelines, this occurs when the joint distribution of inputs and labels differs between the training (source) and application (target) domains, causing source-trained models to underperform on the target. Transfer learning and domain adaptation provide principled ways to mitigate this effect. We study a concrete simulation-to-observation case: semi-supervised domain adaptation (SSDA) to transfer a four-class spectral classifier -- high-redshift quasars, low-redshift quasars, galaxies, and stars -- from J-PAS mock catalogs based on DESI spectra to real J-PAS observations. Our pipeline pretrains on abundant labeled DESIJ-PAS mocks and adapts to the target domain using a small labeled J-PAS subset. We benchmark SSDA against two baselines: a J-PAS--only supervised model trained with the same target-label budget, and a mocks-only model evaluated on held-out J-PAS data. On this held-out J-PAS data, SSDA achieves a macro-F1 score (balancing precision and recall) of and an overall true positive rate of , compared to for the J-PAS--only baseline and for the mocks-only model. The gains are driven primarily by improved quasar classification, especially in the high-redshift subclass ( vs.\ ), yielding better-calibrated candidate lists for spectroscopic targeting (e.g., WEAVE-QSO) and AGN searches. This study shows how modest target supervision enables robust, data-efficient simulation-to-observation transfer when simulations are plentiful but target labels are scarce.
Cite
@article{arxiv.2602.13902,
title = {J-PAS: Semi-Supervised Sim-to-Obs Transfer for Robust Star--Galaxy--Quasar Classification},
author = {Daniel López-Cano and L. Raul Abramo and L. Nakazono and I. Pérez-Ràfols and G. Martínez-Solaeche and J. Chaves-Montero and Matthew M. Pieri and Jailson Alcaniz and Narciso Benitez and Silvia Bonoli and Saulo Carneiro and Javier Cenarro and David Cristóbal-Hornillos and Simone Daflon and Renato Dupke and Alessandro Ederoclite and Rosa González Delgado and Antonio Hernán-Caballero and Carlos Hernández-Monteagudo and Jifeng Liu and Carlos López-Sanjuan and Antonio Marín-Franch and Claudia Mendes de Oliveira and Mariano Moles and Fernando Roig and Laerte Sodré and Keith Taylor and Jesús Varela and Héctor Vázquez Ramió and Jose Vilchez and Javier Zaragoza-Cardiel},
journal= {arXiv preprint arXiv:2602.13902},
year = {2026}
}
Comments
18 pages, 7 figures. To be submitted to ApJ