Lyman Break Galaxy selection and redshift measurement with supervised contrastive learning
๐ arXiv:2608.10080 ยท ๐ฅ PDF ยท 2026-08-10 ยท astro-ph.CO
Authors: J. Choppin de Janvry [arXiv ยท scholar] , C. Yรจche [arXiv ยท scholar] , Arjun Dey [arXiv ยท scholar] , C. Magneville [arXiv ยท scholar] , C. Payerne [arXiv ยท scholar] , J. Aguilar [arXiv ยท scholar] , S. Ahlen [arXiv ยท scholar] , E. Armengaud [arXiv ยท scholar] , S. Bailey [arXiv ยท scholar] , F. Beutler [arXiv ยท scholar] , D. Bianchi [arXiv ยท scholar] , D. Brooks [arXiv ยท scholar] , A. Carnero Rosell [arXiv ยท scholar] , E. Chaussidon [arXiv ยท scholar] , T. Claybaugh [arXiv ยท scholar] , A. Cuceu [arXiv ยท scholar] , K. S. Dawson [arXiv ยท scholar] , A. de la Macorra [arXiv ยท scholar] , A. de Mattia [arXiv ยท scholar] , P. Doel [arXiv ยท scholar] , A. Font-Ribera [arXiv ยท scholar] , J. E. Forero-Romero [arXiv ยท scholar] , E. Gaztaรฑaga [arXiv ยท scholar] , Satya Gontcho A Gontcho [arXiv ยท scholar] , G. Gutierrez [arXiv ยท scholar] , H. K. Herrera-Alcantar [arXiv ยท scholar] , K. Honscheid [arXiv ยท scholar] , M. Ishak [arXiv ยท scholar] , S. Juneau [arXiv ยท scholar] , T. Karim [arXiv ยท scholar] , D. Kirkby [arXiv ยท scholar] , A. Kremin [arXiv ยท scholar] , A. Lambert [arXiv ยท scholar] , M. Landriau [arXiv ยท scholar] , L. Le Guillou [arXiv ยท scholar] , M. Manera [arXiv ยท scholar] , A. Meisner [arXiv ยท scholar] , R. Miquel [arXiv ยท scholar] , S. Nadathur [arXiv ยท scholar] , G. Niz [arXiv ยท scholar] , E. Paillas [arXiv ยท scholar] , N. Palanque-Delabrouille [arXiv ยท scholar] , W. J. Percival [arXiv ยท scholar] , F. Prada [arXiv ยท scholar] , I. Pรฉrez-Rร fols [arXiv ยท scholar] , C. Ravoux [arXiv ยท scholar] , G. Rossi [arXiv ยท scholar] , R. Ruggeri [arXiv ยท scholar] , E. Sanchez [arXiv ยท scholar] , C. Saulder [arXiv ยท scholar] , D. Schlegel [arXiv ยท scholar] , M. Schubnell [arXiv ยท scholar] , H. Seo [arXiv ยท scholar] , J. Silber [arXiv ยท scholar] , M. Siudek [arXiv ยท scholar] , G. Tarlรฉ [arXiv ยท scholar] , B. A. Weaver [arXiv ยท scholar] , H. Zou [arXiv ยท scholar]
๐ Abstract
Some of the next steps for high-precision cosmology lie within the high-redshift, high-density universe. Spectroscopic survey experiments such as the Dark Energy Spectroscopic Instrument (DESI)'s second phase DESI Run 2 will shift towards probing Lyman Break Galaxy (LBG) populations from z$\sim$2 to z$\sim$4.5. For this faint sample, spectroscopic redshift measurement and sample decontamination remains a challenge, even after target selection. We propose an approach based on supervised weighted contrastive learning, in order to both learn a redshift representation for spectra and decontaminate the sample from quasars and low redshift emission line galaxies. This strategy generalizes the contrastive learning loss approach with continuous relationship weights, such that the network simultaneously learns redshift and classification tasks. The model shows stronger outlier classification and comparable redshift identification performances when compared to the previous network used for DESI (a modified version of QuasarNET) on the same dataset. In particular, contrastive learning is well suited to the small, visually-inspected sample used for training and testing, especially given the multi-task nature of this work.