Multiwavelength Spectral Analysis and Neural Network Classification of Counterparts to 4FGL Unassociated Sources

Abstract

The Fermi-LAT unassociated sources represent some of the most enigmatic gamma-ray sources in the sky. Observations with the Swift-XRT and -UVOT telescopes have identified hundreds of likely X-ray and UV/optical counterparts in the uncertainty ellipses of the unassociated sources. In this work we present spectral fitting results for 205 possible X-ray/UV/optical counterparts to 4FGL unassociated targets. Assuming that the unassociated sources contain mostly pulsars and blazars, we develop a neural network classifier approach that applies gamma-ray, X-ray, and UV/optical spectral parameters to yield a descriptive classification of unassociated spectra into pulsars and blazars. From our primary sample of 174 Fermi sources with a single X-ray/UV/optical counterpart, we present 132 P bzr > 0.99 likely blazars and 14 P bzr < 0.01 likely pulsars, with 28 remaining ambiguous. These subsets of the unassociated sources suggest a systematic expansion to catalogs of gamma-ray pulsars and blazars. Compared to previous classification approaches our neural network classifier achieves significantly higher validation accuracy and returns more bifurcated P bzr values, suggesting that multiwavelength analysis is a valuable tool for confident classification of Fermi unassociated sources.

Using Neural Networks to Differentiate Newly Discovered BL Lacertae Objects and FSRQs among the 4FGL Unassociated Sources Employing Gamma-Ray, X-Ray, UV/Optical, and IR Data

“…This work identifies and classifies four FSRQs and 50 BL Lac objects from the 112 blazar candidates identified through X-ray counterpart searches among the 4FGL-DR3 unassociated sources by Kerby et al (2021). In addition, upon using SIMBAD and SDSS catalogs to conduct further searches for any known sources at the UV/optical positions of these sources, we found more BL Lac objects, FSRQs, radio galaxies, as well as a cataclysmic variable star (CV) and two Seyfert type I galaxies.…”

“…In the past few years, various authors explored using machine learning to classify blazars such as Doert & Errando (2014), Salvetti et al (2017), Kang et al (2019), , Kovačević et al (2020), andChiaro et al (2021). Most recently, Kerby et al (2021) (K21 from now on) has found blazar as well as pulsar candidates among these 238 X-ray sources employing the gamma-ray, X-ray, and UV/ optical properties using a method of neural networks. The objective of this paper is to classify these blazar candidates further into their subclasses: FSRQs and BL Lac objects using a neural network classifier employing the data from gammaray, X-ray, UV/optical, and IR regimes.…”

Identifying the 3FHL Catalog. VI. Swift Observations of 3FHL Unassociated Objects with Source Classification via Machine Learning

“…Just as we used WISE data to create multiple new features, one can combine other multiwavelength properties into an original feature to be used for a machine-learning algorithm. Kerby et al (2021a) used the logarithmic ratio between X-ray and gamma-ray flux and found this to be a significant feature, which was used in a subsequent paper (Kerby et al 2021b). Kerby et al (2021b) also notes that neural networks tend to perform better than random forest algorithms, with increased confidence and a reduction in the number of ambiguous sources.…”

“…Kerby et al (2021a) used the logarithmic ratio between X-ray and gamma-ray flux and found this to be a significant feature, which was used in a subsequent paper (Kerby et al 2021b). Kerby et al (2021b) also notes that neural networks tend to perform better than random forest algorithms, with increased confidence and a reduction in the number of ambiguous sources.…”

Classification of Fermi-LAT blazars with Bayesian neural networks

“…Machine learning methods, including neural networks, have been used for the classification of γ-ray sources in various analyses, ranging from the identification of AGN and pulsar candidates [11][12][13][14][15][16] and of blazars [17][18][19][20][21] to the search for new exotic source classes such as dark matter subhalos [22]. However, apart from the usual performance tests done on the training and testing data sets, it is not clear in general how to estimate the uncertainty associated with the machine learning output.…”