The complete catalog of ML-synthesized and spectroscopically measured redshifts for 1.5 million Gaia CRF3 quasars

Supplemental dataset for the research paper "Kinematic Distortions of the High-Redshift Universe as Seen from Quasar Proper Motions" in Nature Astronomy, 2025 The catalog includes Machine-Learning generated and spectroscopically measured (when available) redshifts for 1.5 million Gaia CRF3 (DR3) optical sources, mostly quasars. For this version, supervised and trained neural network method has been used. The ML-prediction is based on several classifiers extracted from the unWISE catalog of MIR photometry and the Gaia DR3 main source catalog. For the network training, the spectroscopic redshifts from SDSS have been used. The columns are 1) Gaia source id, 2) RA (2016) in deg, 3) Dec (2016) in deg, 4) ML-predicted redshift, 5) SDSS mesured redshift (-9 if not available).