Improved photometric redshift estimations through self-organising map-based data augmentation 2025 Improved photometric redshift estimations through self-organising map-based data augmentation – References

Yun-Hao Zhang (张云皓), Joe Zuntz, Irene Moskowitz, Eric Gawiser, Konrad Kuijken, Marika Asgari, Henk Hoekstra, Alex I. Malz, Ziang Yan (颜子昂), Tianqing Zhang, and The LSST Dark Energy Science Collaboration · 2025

Read Improved photometric redshift estimations through self-organising map-based data augmentation 2025 Improved photometric redshift estimations through self-organising map-based data augmentation – References by Yun-Hao Zhang (张云皓), Joe Zuntz, Irene Moskowitz, Eric Gawiser, Konrad Kuijken, Marika Asgari, Henk Hoekstra, Alex I. Malz, Ziang Yan (颜子昂), Tianqing Zhang, and The LSST Dark Energy Science Collaboration (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-20903.txt.

Introduction

Yun-Hao Zhang and a large team of co-authors present a method in two thousand twenty-five to improve galaxy distance estimates. Photometric redshifts infer a galaxy's distance from its colors, using a training set with known distances. Some regions of color space are thinly sampled by that training set, which hurts accuracy.

The method projects galaxy spectra onto a self-organising map, a two-dimensional layout of similar cases. Sparse regions are then filled with simulated galaxies from an independent catalog. Tests used mock data modeled on the coming Legacy Survey of Space and Time. Added galaxies cut biases and reduced failures for distant galaxies by up to about half. The gain matters for cosmology with the Rubin Observatory.

Catalog record

CREATORYun-Hao Zhang (张云皓), Joe Zuntz, Irene Moskowitz, Eric Gawiser, Konrad Kuijken, Marika Asgari, Henk Hoekstra, Alex I. Malz, Ziang Yan (颜子昂), Tianqing Zhang, and The LSST Dark Energy Science Collaboration
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE86 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-20903

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Improved photometric redshift estimations through self-organising map-based data augmentation 2025 Improved photometric redshift estimations through self-organising map-based data augmentation – References Yun-Hao Zhang (张云皓), Joe Zuntz, Irene Moskowitz, Eric Gawiser, Konrad Kuijken, Marika Asgari, Henk Hoekstra, Alex I. Malz, Ziang Yan (颜子昂), Tianqing Zhang, and The LSST Dark Energy Science Collaboration Abstract We introduce a framework for the enhanced estimation of photometric redshifts using Self-Organising Maps (SOMs). Our method projects galaxy Spectral Energy Distributions (SEDs) onto a two-dimensional map, identifying regions that are sparsely sampled by existing spectroscopic observations. These under-sampled areas are then augmented with simulated galaxies, yielding a more representative spectroscopic training dataset. To assess the efficacy of this SOM-based data augmentation in the context of the forthcoming Legacy Survey of Space and Time (LSST), we employ mock galaxy catalogues from the OpenUniverse2024 project and generate synthetic datasets that mimic the expected photometric selections of LSST after one (Y1) and ten (Y10) years of observation. We construct 501 degraded realisations by sampling galaxy colours, magnitudes, redshifts and spectroscopic success rates, in order to emulate the compilation of a wide array of realistic spectroscopic surveys. Augmenting the degraded mock datasets with simulated galaxies from the independent CosmoDC2 catalogues has markedly improved the performance of our photometric redshift estimates compared to models lacking this augmentation, particularly for high-redshift galaxies ( z true ≳ 1.5 ) (z_{\mathrm{true}}\gtrsim 1.5) . This improvement is manifested in notably reduced systematic biases and a decrease in catastrophic failures by up to approximately a factor of 2 2 , along with a reduction in information loss in the conditional density estimations. These results underscore the effectiveness of SOM-based augmentation in refining photometric redshift estimation, thereby enabling more robust analyses in cosmology and astrophysics for the NSF-DOE Vera C. Rubin Observatory. 1 Introduction Cosmological redshift underpins myriad applications in modern astronomy. Accurate redshift estimates permit the conversion of observed magnitudes of distant galaxies into intrinsic luminosities, allowing studies of the galaxy luminosity function ( Schechter 1976 ; Finkelstein et al. 2015 ; Bouwens et al. 2023 , e.g.) , the evolution of galaxy size and morphology across cosmic epochs ( Conselice 2014 ; Shibuya et al. 2015 , e.g.) , and scaling relations between galaxy stellar masses and properties of their interstellar medium ( Kennicutt 1998 ; Tremonti et al. 2004 , e.g.) , central supermassive black holes ( Kormendy & Ho 2013 , e.g.) , and host haloes ( Behroozi et al. 2013 ; Wechsler & Tinker 2018 , e.g.) . In cosmology, Type Ia supernova host galaxy redshifts provided the first direct evidence of the accelerating expansion of the Universe ( Riess et al. 1998 ; Perlmutter et al. 1999 ) . Baryon Acoustic Oscillations (BAOs), measured from galaxy clustering, act as a standard ruler for cosmic distances ( Eisenstein et al. 2005 ) , while large-scale structure surveys map the growth of density fluctuations ( Peebles 1980 ; Springel et al. 2006 ) . The weak gravitational lensing of distant galaxies ( Tyson et al. 1990 ; Kaiser 1998 ; Bartelmann & Schneider 2001 ) enables the reconstruction of the matter distribution and serves as a sensitive probe of dark energy through cosmic-shear tomography ( Kilbinger 2015 ) . Collectively, these observables demonstrate how cosmological redshift is the keystone linking local galaxy properties to the geometry and dynamics of the Universe, and a critical underpinning of the concordance \Lambda CDM model. Recent wide-field imaging surveys offer multi-wavelength photometry for hundreds of millions of galaxies, such as the Kilo-Degree Survey ( Wright et al. 2024 , KiDS 1 1 1 https://kids.strw.leidenuniv.nl ;) , the Dark Energy Survey ( Bechtol et al. 2025 , DES 2 2 2 https://www.darkenergysurvey.org ;) , and the Hyper Suprime-Cam Survey ( Li et al. 2022 , HSC 3 3 3 https://hsc.mtk.nao.ac.jp/ssp/ ;) . Nonetheless, acquiring precise redshifts using spectroscopy for such extensive samples remains exceedingly resource-demanding. As a result, studies on large-scale structures and galaxy evolution are increasingly dependent on photometric redshift methods, which estimate redshifts from broadband flux observations, a technique initially proposed by Koo 1985 ; Loh & Spillar 1986 . Next‐generation facilities – including the NSF-DOE Vera C. Rubin Observatory ( LSST Science Collaboration et al. 2009 , Rubin 4 4 4 https://rubinobservatory.org ;) , the Euclid mission ( Euclid Collaboration et al. 2025 , Euclid 5 5 5 https://www.cosmos.esa.int/web/euclid ;) , the Nancy Grace Roman Space Telescope ( Spergel et al. 2015 , Roman 6 6 6 https://roman.gsfc.nasa.gov ;) and the China Space Station Telescope ( Gong et al. 2019 , CSST 7 7 7 http://www.bao.ac.cn/csst/ ;) – will deliver unprecedented imaging depth and sky coverage. Exploiting their full cosmological potential to probe dark matter and dark energy, however, demands exceptional control of systematic uncertainties, especially in photometric redshift performance ( Mandelbaum 2018 ) . In recent decades, numerous techniques for estimating photometric redshifts have been developed and are generally classified into two categories based on the type of data and the priors used during inference, as reviewed in Newman & Gruen 2022 . Comprehensive descriptions of most of these techniques are available through the Redshift Assessment Infrastructure Layers ( The RAIL Team et al. 2025 , RAIL 8 8 8 https://github.com/LSSTDESC/rail ;) platform for Legacy Survey of Space and Time (LSST) Dark Energy Science Collaboration (DESC). Specifically, the most traditional approach, template fitting, matches observed broadband photometry to libraries of galaxy Spectral Energy Distribution (SED) templates, using software packages such as LePhare ( Arnouts et al. 1999 ; Ilbert et al. 2006 ) , BPZ ( Benítez 2000 ; Coe et al. 2006 ) , ZEBRA ( Feldmann et al. 2006 ) , EAZY ( Brammer et al. 2008 ) , and DELIGHT ( Leistedt & Hogg 2017 ) . By incorporating stellar population synthesis models, template fitting can recover redshift posteriors from noisy magnitude measurements; however, the limited completeness of template sets and potential biases in the adopted priors can introduce systematic errors and constrain the ultimate accuracy of the estimates ( Hildebrandt et al. 2010 ; Dahlen et al. 2013 , e.g.) . Conversely, a variety of machine-learning algorithms have been devised to characterise the complex relationship between galaxy broadband SEDs and redshifts, utilising galaxies with known redshifts. These include artificial neural networks ( Collister & Lahav 2004 ; Sadeh et al. 2016 , ANNz ;) , decision trees with random forest techniques, Gaussian processes ( Almosallam et al. 2016 , GPz ;) , the quasi-Newton algorithm ( Cavuoti et al. 2012 , MLPQNA ;) , hybrid density estimation ( Cavuoti et al. 2017 , METAPHOR ;) , boosted decision trees combined with basis function decomposition ( Izbicki & Lee 2017 ; Dalmasso et al. 2020 ; Schmidt et al. 2020 , FlexZBoost ;) , the normalising flow algorithm ( Crenshaw et al. 2024 , PZFlow ; ) , and stratified learning ( Revsbech et al. 2018 ; Autenrieth et al. 2024 ; Moretti et al. 2025 , StratLearn-z ;) , and These data-driven approaches rely on training samples with high completeness and precisely measured redshifts, typically drawn from wide‐area spectroscopic surveys such as the Sloan Digital Sky Survey ( York et al. 2000 , SDSS 9 9 9 https://www.sdss.org ;) , the

Read the complete text