Astroinformatics of galaxies and quasars: a new general method for photometric redshifts estimation

Abstract

With the availability of the huge amounts of data produced by current and future large multiband photometric surveys, photometric redshifts have become a crucial tool for extragalactic astronomy and cosmology. In this paper we present a novel method, called Weak Gated Experts (WGE), which allows us to derive photometric redshifts through a combination of data mining techniques. The WGE, like many other machine learning techniques, is based on the exploitation of a spectroscopic knowledge base composed by sources for which a spectroscopic value of the redshift is available. This method achieves a variance σ2(Δz) = 2.3 × 10−4 [σ2(Δz) = 0.08, where Δz=zphot−zspec] for the reconstruction of the photometric redshifts for the optical galaxies from the Sloan Digital Sky Survey (SDSS) and for the optical quasars, respectively, while the root mean square (rms) of the Δz variable distributions for the two experiments is, respectively, equal to 0.021 and 0.35. The WGE provides also a mechanism for the estimation of the accuracy of each photometric redshift. We also present and discuss the catalogues obtained for the optical SDSS galaxies, for the optical candidate quasars extracted from the Data Release 7 of SDSS photometric data set (the sample of SDSS sources on which the accuracy of the reconstruction has been assessed is composed of bright sources, for a subset of which spectroscopic redshifts have been measured) and for optical SDSS candidate quasars observed by GALEX in the ultraviolet range. The WGE method exploits the new technological paradigm provided by the virtual observatory and the emerging field of astroinformatics.

A catalogue of photometric redshifts for the SDSS-DR9 galaxies

“…In the upper half we provide standard statistical indicators (see text for an explanation) used to evaluate the performances of photo-z methods. We also include the same indicators for Laurino et al (2011). In the lower half of the table, we report the fraction (in percentage) of outliers computed using a fixed threshold of 0.15, the more meaningful 1, and 2σ clipping thresholds and the values of skewness and kurtosis of the σ (Δz norm ) distributions.…”

“…The redshift range starting from z = 0 up to the redshift z 1 = 0.115 includes 50% of the objects in the test set; the second redshift bin, using the range from z 1 = 0.115 to z 2 = 0.177, corresponds to an additional 25% of the sample; the third bin, using the range z 2 = 0.177 to z 3 = 0.345, corresponds to an additional 15% and, finally, the fourth bin includes all remaining objects (redshift >z 3 = 0.345). The last row reports the statistics in the redshift range [0.05, 0.6] corresponding to the same range covered in Laurino et al (2011). 1.8…”

“…We also derived statistics in the redshift range [0.05, 0.6] corresponding to the same range covered in Laurino et al (2011) (in order to allow for a fair comparison).…”

Multi-wavelength selection and identification of gamma-ray blazar candidates

“…Applying this method, Cowperthwaite et al (2013) recently identified thirteen gamma-ray emitting blazar candidates from a sample of 102 previously unidentified sources selected from Astronomer's Telegrams and the literature. Using the X-ray emission in place of the 22 μm detection, Paggi et al (2013) proposed a method to select γ-ray blazar candidates among Swift-XRT sources considering those that feature a WISE counterpart detected at least in the first 3 bands, and with IR colors compatible with the 90% two-dimensional densities of known γ-ray blazar evaluated using the Kernel Density Estimation (KDE) technique (Richards et al 2004;D'Abrusco, Longo & Walton 2009;Laurino et al 2011 and reference therein), so selecting 37 new γ-ray blazar candidates (Fig. 2, left panel).…”

Determining spectroscopic redshifts by usingknearest neighbor regression

“…There are many applications of machine-learning techniques in astronomy (see Borne 2009;Ball & Brunner 2010). So far, spectroscopically derived properties have mainly A&A 576, A132 (2015) been used as ground truth to estimate redshifts on photometric data, for example in Laurino et al (2011), Gieseke et al (2011), Polsterer et al (2013). In contrast less attention has been paid to the application of machine learning to the spectral data itself (see Richards et al 2009;Meusinger et al 2012), which can be mainly attributed to the "curse of dimensionality" (see Bellman & Bellman 1961).…”