QSO Selection and Photometric Redshifts with Neural Networks

TL;DR

This paper presents a neural network-based method for selecting quasars and estimating their redshifts using SDSS photometry, achieving high stellar rejection and accurate redshift predictions for cosmological surveys.

Contribution

It introduces a multilayer perceptron neural network for quasar selection and photometric redshift estimation, improving efficiency and accuracy over previous methods.

Findings

Achieves 99.6% stellar rejection at 50% quasar efficiency

Photometric redshift precision of about 0.1 for BAO studies

Enables effective quasar target selection for large cosmological surveys

Abstract

Baryonic Acoustic Oscillations (BAO) and their effects on the matter power spectrum can be studied by using the Lyman-alpha absorption signature of the matter density field along quasar (QSO) lines of sight. A measurement sufficiently accurate to provide useful cosmological constraints requires the observation of ~100000 quasars in the redshift range 2.2<z<3.5 over ~8000 deg2. Such a survey is planned by the Baryon Oscillation Spectroscopic Survey (BOSS) project of the Sloan Digital Sky Survey (SDSS-III).In practice, one needs a stellar rejection of more than two orders of magnitude with a selection efficiency for quasars better than 50% up to magnitudes as large as g ~ 22. To obtain an appropriate target list and estimate quasar redshifts, we have developed an Artificial Neural Networks (NN) with a multilayer perceptron architecture. The input variables are photometric measurements, i.e. the object magnitudes and their errors in the five bands (ugriz) of the SDSS photometry. For target selection, we achieve a non-quasar point-like object rejection of 99.6% and 98.5% for a quasar efficiency of, respectively, 50% and 85%. The photometric redshift precision is of the order of 0.1 over the region relevant for BAO studies.