Catalog of quasars from the Kilo-Degree Survey Data Release 3

Abstract

We present a catalog of quasars selected from broad-band photometric ugri data of the Kilo-Degree Survey Data Release 3 (KiDS DR3). The QSOs are identified by the random forest (RF) supervised machine learning model, trained on Sloan Digital Sky Survey (SDSS) DR14 spectroscopic data. We first cleaned the input KiDS data of entries with excessively noisy, missing or otherwise problematic measurements. Applying a feature importance analysis, we then tune the algorithm and identify in the KiDS multiband catalog the 17 most useful features for the classification, namely magnitudes, colors, magnitude ratios, and the stellarity index. We used the t-SNE algorithm to map the multidimensional photometric data onto 2D planes and compare the coverage of the training and inference sets. We limited the inference set to r < 22 to avoid extrapolation beyond the feature space covered by training, as the SDSS spectroscopic sample is considerably shallower than KiDS. This gives 3.4 million objects in the final inference sample, from which the random forest identified 190,000 quasar candidates. Accuracy of 97% (percentage of correctly classified objects), purity of 91% (percentage of true quasars within the objects classified as such), and completeness of 87% (detection ratio of all true quasars), as derived from a test set extracted from SDSS and not used in the training, are confirmed by comparison with external spectroscopic and photometric QSO catalogs overlapping with the KiDS footprint. The robustness of our results is strengthened by number counts of the quasar candidates in the r band, as well as by their mid-infrared colors available from the Wide-field Infrared Survey Explorer (WISE). An analysis of parallaxes and proper motions of our QSO candidates found also in Gaia DR2 suggests that a probability cut of p QSO > 0.8 is optimal for purity, whereas p QSO > 0.7 is preferable for better completeness. Our study presents the first comprehensive quasar selection from deep high-quality KiDS data and will serve as the basis for versatile studies of the QSO population detected by this survey. We publicly release the resulting catalog at http://kids.strw.leidenuniv.nl/DR3/quasarcatalog.php, and the code at https://github.com/snakoneczny/kids-quasars.

Unsupervised star, galaxy, QSO classification

“…It is clear that the verification procedure of Pâris et al (2018) has removed a lot of interlopers, particularly sources with starlike colours. Nakoneczny et al (2019) used KIDS DR3 to search for quasars using ug r i colours and magnitudes, and the stellarity morphological index. They created a parent catalogue of 3.4 10 6 sources and using SDSS DR14 spectroscopic labels, they identified a total of 190,000 quasar candidates (r <22).…”

“…Out of the 837,624 sources in common, we examined the overlap of quasar positive classifications. We selected KIDS-DR3 quasars using a threshold of Pr[QSO]>0.7 suggested for optimized completeness in Nakoneczny et al (2019). Using the matched sample, we find that the KIDS-DR3 catalogue identifies 29,878 quasar candidates while the KiDS-VW sam- ple identifies 40,621 quasars.…”

“…According to the feature importance ranking discussed in Nakoneczny et al (2019), the stellarity index carried the highest weight. Hence, we conclude that the ug r i colours where suffiencent to disentangle between stars and quasars leading to the creation of a pure quasar catalogue.…”

“…In §6 we discuss the impact of the input attributes, including colours and half light radii, as well as the photometric depth and photometric uncertainties. In §8 we compare our results to the automatic labels of the Sloan Digital Sky Survey (SDSS DR14), the refined labels of the SDSS Quasars catalogue of Pâris et al (2018), and the KIDS DR3 quasar catalogue of Nakoneczny et al (2019). Finally, we apply and publicly release our classifications on the Kilo Degree Survey, matched with the VIKING and ALLWISE surveys (∼3 million sources).…”

Surveying the reach and maturity of machine learning and artificial intelligence in astronomy

“…New methods are being developed to find anomalous objects in astronomical datasets, such as the work by Baron and Poznanski () using an unsupervised RF to find outliers among SDSS galaxies. Promising avenues involve a combination of unsupervised learning methods, such as isolation forests (Liu, Ting, & Zhou, ); dimensionality reduction, such as PCA, t‐Distributed Stochastic Neighbor Embedding (van der Maaten & Hinton, ), for example, Reis, Poznanski, Baron, Zasowski, and Shahaf () and Nakoneczny et al (), self‐organizing maps (SOMs; Kohonen, ), for example, Carrasco Kind and Brunner () and Armstrong et al (), or the latent space of a variational encoder (e.g., Ma et al, ; Yang & Li, ). Novel visualization techniques are also contributing.…”