Evaluating the Classification of Fermi BCUs from the 4FGL Catalog Using Machine Learning

Abstract

The recently published fourth Fermi Large Area Telescope source catalog (4FGL) a)b) , reports 5065 gamma-ray sources in terms of direct observational gamma-ray properties. Among the sources, the largest population is the Active Galactic Nuclei (AGN), which consists of 3137 blazars, 42 radio galaxies, and 28 other AGNs. The blazar sample comprises 694 flat-spectrum radio quasars (FSRQs), 1131 BL Lac-type objects (BL Lacs), and 1312 blazar candidates of an unknown type (BCUs). The classification of blazars is difficult using optical spectroscopy given the limited knowledge with respect to their intrinsic properties, and the limited availability of astronomical observations. To overcome these challenges, machine learning algorithms are being investigated as alternative approaches. Using the 4FGL catalog, a sample of 3137 Fermi blazars with 23 parameters is systematically selected. Three established supervised machine learning algorithms (random forests (RFs), support vector machines (SVMs), artificial neural networks (ANNs)) are employed to general predictive models to classify the BCUs. We analyze the results for all of the different combinations of parameters. Interestingly, a previously reported trend the use of more parameters leading to higher accuracy is not found. Considering the least number of parameters used, combinations of eight, 12 or 10 parameters in the SVM, ANN, or RF generated models achieve the highest accuracy (Accuracy ≃ 91.8%, or ≃ 92.9%). Using the combined classification results from the optimal combinations of parameters, 724 BL Lac type candidates and 332 FSRQ type candidates are predicted; however, 256 remain without a clear prediction.

Searching for AGN and Pulsar Candidates in 4FGL Unassociated Sources Using Machine Learning

“…For example, the source numbered as 4FGL J0531.7+1241c is obtained as uncertain type, while it is evaluated as an AGN in ANN classifier and evaluated as an other γ-ray source in RF classifier. Hence, the accuracy is improved, although the number of candidates is reduced (e.g., Kang et al 2019b). The combined test results of the two algorithms are shown in Table 4.…”

“…Since different features play different roles in the classifiers, the selection of suitable input features for the SML is necessary. Noticing that, i) More input features do not always result in higher accuracy (Kang et al 2019b); ii) More features need more computation; iii) The favorable features for the selection of the AGNs are different from those for pulsars, we further select the features for the two steps from the 36 usable features.…”

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

“…• Active galactic nuclei and quasars. A common theme in this field is the need for classification and detection methods, including assigning morphological types to radio-detected active galactic nuclei with a CNN (Ma et al, 2019), identifying blazar candidates in the Fermi-LAT (3LAC) Clean Sample (Kang et al, 2019), detecting rare high-redshift, extremely luminous quasars (Schindler et al, 2017), and discriminating populations of broad absorption line quasars (BALQs) from non-BALQs in SDSS data releases (Yong et al, 2018). • Cosmological simulations.…”

“…Duev et al () trained a CNN to discover fast‐moving candidates from ZTF observations in order to more reliably identify potentially hazardous near‐Earth objects. Active galactic nuclei and quasars . A common theme in this field is the need for classification and detection methods, including assigning morphological types to radio‐detected active galactic nuclei with a CNN (Ma et al, ), identifying blazar candidates in the Fermi‐LAT (3LAC) Clean Sample (Kang et al, ), detecting rare high‐redshift, extremely luminous quasars (Schindler et al, ), and discriminating populations of broad absorption line quasars (BALQs) from non‐BALQs in SDSS data releases (Yong et al, ). Cosmological simulations . ML is providing new methods for examining the outputs of cosmological simulations, leading to new insights about the connections between physical properties of galaxies, dark matter halos and the cosmic environment.…”

An Empirical “High-confidence” Candidate Zone for Fermi BL Lacertae Objects

“…Indeed, some potential BL Lac or FSRQ candidates can be identified from the BCUs sample in the 2FGL/3FGL catalogues using different approaches such as supervised machine learning (e.g., support vector machine [SVM] and random forest [RF]; Hassan et al (2013)), neural network (Chiaro et al 2016), artificial neural network (ANN; Salvetti et al 2017), multivariate classification method (Lefaucheur & Pita 2017), and by statisical analysis of the broadband spectral properties (including spectral indices in the gamma-ray, X-ray, optical, and radio bands; Yi et al 2017). In addition, we've identified potential BL Lacs and FS-RQs candidates from the 3LAC Clean sample using 4 different SML algorithms (Mclust Gaussian finite mixture models, Decision trees, RF, and SVM; Kang et al 2019a [Paper I]) and from the 4FGL catalogue using 3 different SML algorithms (ANN, RF, and SVM; Kang et al 2019b). Nevertheless, the final confirmation of the BCU nature of candidates in all above approaches is subject to the observations of optical spectroscopy or counterparts in other wavelength (e.g., Massaro et al 2014;Álvarez Crespo et al 2016a,b,c;Massaro et al 2016;Marchesini et al 2016;Klindt et al 2017;Peña-Herazo et al 2017;Marchesi et al 2018;Desai et al 2019;Marchesini et al 2019;Peña-Herazo et al 2019), or broadband spectral features (e.g., Fermi/LAT collaboration, Massaro et al 2009Massaro et al , 2012Massaro et al , 2016Álvarez Crespo et al 2016a,b,c).…”