Machine Learning and Data-Driven Discovery in Large Extragalactic Surveys

Extragalactic astronomy entered a data-rich era in which Euclid, the Vera C. Rubin Observatory’s Legacy Survey of Space and Time, Roman, 4MOST, and DESI will produce unprecedented imaging, spectroscopic, and time-domain datasets. Machine learning is increasingly used for galaxy classification, photometric redshift estimation, active galactic nucleus identification, strong-lensing detection, anomaly discovery, and physical inference, just to name a few applications. Despite major advances, challenges remain in handling heterogeneous noise, selection effects, missing data, limited labels, simulation–observation differences, and model interpretability. Further research is needed to ensure that these methods provide robust physical insight rather than only improved performance metrics.This Research Topic aims to advance reliable, interpretable, and physically informative machine-learning approaches for extragalactic astronomy. It will explore how data-driven methods can scale survey analysis, discover rare or unexpected objects, connect simulations with observations, and improve understanding of galaxy formation and evolution. Studies addressing uncertainty, model transferability, cross-survey validation, and the physical meaning of learned representations will be particularly welcome.To gather further insights into machine learning and data-driven discovery in large extragalactic surveys, we welcome contributions focused on galaxies, active galactic nuclei, gravitational lensing, and large-scale structure, including:• Galaxy morphology, spectral classification, and photometric redshift estimation.• Active galactic nucleus identification, variability, and outflow prediction.• Supervised, unsupervised, self-supervised, and phylogenetic approaches to galaxy evolution.• Strong-lensing detection, anomaly identification, and rare-object discovery.• Simulation-based inference, neural emulators, and surrogate models.• Representation learning for heterogeneous survey data.• Uncertainty quantification, domain adaptation, and missing-data methods.• Interpretability, robustness, reproducibility, and benchmarking of astrophysical models.• Data-driven discovery of new physical relationships and extragalactic populations. Topic editors: Massimo Brescia, Mara Salvato Source: https://www.frontiersin.org/research-topics/86698/machine-learning-and-data-driven-discovery-in-large-extragalactic-surveys