FOMO-Z. Optimal galaxy classification, photometric redshifts, and reconstruction of population distributions

Abstract

We present FOMO-Z (Flexible Optimizer of Multi-dimensional Operators and photo-$z$s), a set of statistical methods implemented in Python for galaxy classification and probabilistic inference from photometry. The methods use a reference sample with known galaxy properties. They rely on explicit, physically interpretable criteria and make no assumptions about the functional form of the color distributions. Binary selections are calibrated using either linear color conditions, particularly suited to background galaxy selection, or flexible color--color grids, seeking to maximize completeness given a threshold on purity. For probabilistic inference, the code identifies reference objects with compatible colors within photometric uncertainties to estimate each test galaxy's probability of satisfying user-defined criteria, as well as probability density functions of continuous properties such as redshift. The code also reconstructs property distributions of selected galaxy populations using photometrically similar galaxies in the reference sample. This framework supports prior weights and reference redshift probability distributions. A built-in feature-selection method identifying informative colors is also introduced. We demonstrate applications to background galaxy selection, galaxy population classification, tomographic redshift calibration, and photometric redshift estimation. The code achieves results consistent with those obtained through machine-learning approaches, such as self-organizing maps, while remaining computationally efficient, making it suitable for large photometric surveys and future Stage IV cosmological analyses. The code is publicly available at https://github.com/GiorgioLesci/FOMO-Z.