Determination of Galaxy Photometric Redshifts Using Conditional Generative Adversarial Networks (CGANs)

Abstract

Accurate and reliable galaxy redshift determination is a key requirement for wide-field photometric surveys. The estimation of pho-tometric redshifts for galaxies has traditionally been addressed using artificial intelligence techniques trained on calibration samples, where both photometric and spectroscopic data are available. In this paper, we present the first algorithmic approach for photometric redshift estimation using Conditional Generative Adversarial Networks (CGANs). The pro-posed implementation is capable of producing both point estimates and probability density functions for photometric redshifts. The methodology is tested on Year 1 data from the Dark Energy Survey (DES-Y1) and compared against the current state-of-the-art Mixture Density Network (MDN) algorithm. The CGAN approach achieves comparable quality metrics to the MDN, demonstrating its potential and opening the door to the use of adversarial networks in photometric redshift estimation.