They trained an algorithm to identify celestial objects in the survey by feeding them with spectroscopic measurements that provide specific object classifications and distances. Lead study author Robert Beck, former IfA postdoctoral cosmology fellow, said: “By using a sophisticated optimization algorithm, we took advantage of the Spectral Training Suite when Nearly 4 million light sources teach the neural network to predict source types and distances galaxies, while simultaneously correcting light extinction by dust in the Milky Way. ”
This enabled the neural network to achieve a classification accuracy of 98.1% for galaxies, 97.8% for stars, and 96.6% for quasars, and the team says their catalog is twice the size of the largest pre-universe map, created by the Sloan Digital Sky Survey (SDSS). It covers a third of the sky.
A smaller initial version of the catalog has already helped discover the largest void in the universe – a possible cause of the cold spot, which is a vast and mysteriously cold area in the sky.The entire 300GB database can now be downloaded as a computer-readable table.