DeepDISC-photoz: Deep Learning-Based Photometric Redshift Estimation for Rubin LSST

• DOI: 10.33232/001c.136809
• Publication Date: 2025-04-16T07:59:36Z
• Authors: Grant Merz; Xin Liu; Samuel Schmidt; Alex I. Malz; Tianqing Zhang; Doug Branton; Colin J. Burke; Melissa Delucchi; Yaswant Sai Ejjagiri; Jeremy Kubica; Yichen Liu; Olivia Lynn; Drew Oldag; The LSST Dark Ene...
• Keywords: Photometric redshift

Photometric redshifts will be a key data product for the Rubin Observatory Legacy Survey of Space and Time (LSST) as well future ground space-based surveys. The need photometric redshifts, or photo-zs, arises from sparse spectroscopic coverage observed galaxies. LSST is expected to observe billions objects, making it crucial have photo-z estimator that accurate efficient. To end, we present DeepDISC photo-z, an extension framework. base network simultaneously detects, segments, classifies objects in multi-band coadded images. We introduce capabilities by adding redshift estimation Region Interest head, which produces probability distribution function each detected object. On simulated images, outperforms traditional catalog-based estimators, both point estimate probabilistic metrics. validate examining dependencies on systematics including galactic extinction, blending PSF effects. also examine impact quality size training set model. find biggest factor signal-to-noise imaging data, see reduction scatter approximately proportional image signal-to-noise. Our code fully public integrated RAIL package ease use comparison other codes at https://github.com/LSSTDESC/rail_deepdisc