Classification of Fermi-LAT sources with deep learning using energy and time spectra

Abstract

Despite the growing number of gamma-ray sources detected by the Fermi-Large Area Telescope (LAT), about one third of the sources in each survey remains of uncertain type. We present a new deep neural network approach for the classification of unidentified or unassociated gamma-ray sources in the last release of the Fermi-LAT catalogue (4FGL-DR2) obtained with 10 years of data. In contrast to previous work, our method directly uses the measurements of the photon energy spectrum and time series as input for the classification, instead of specific, human-crafted features. Dense neural networks, and for the first time in the context of gamma-ray source classification recurrent neural networks, are studied in depth. We focus on the separation between extragalactic sources, i.e. Active Galactic Nuclei, and Galactic pulsars, and on the further classification of pulsars into young and millisecond pulsars. Our neural network architectures provide powerful classifiers, with a performance that is comparable to previous analyses based on human-crafted features. Our benchmark neural network predicts that of the sources of uncertain type in the 4FGL-DR2 catalogue, 1050 are Active Galactic Nuclei and 78 are Galactic pulsars, with both classes following the expected sky distribution and the clustering in the variability-curvature plane. We investigate the problem of sample selection bias by testing our architectures against a cross-match test data set using an older catalogue, and propose a feature selection algorithm using autoencoders. Our list of high-confidence candidate sources labelled by the neural networks provides a set of targets for further multiwavelength observations addressed to identify their nature. The deep neural network architectures we develop can be easily extended to include specific features, as well as multiwavelength data on the source photon energy and time spectra coming from different instruments.

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.…”

“…We use neural networks trained on the energy spectra of known BLL and FSRQ to classify blazars of uncertain type. Using the energy dependent flux instead of derived features as input for neural network classifiers has already been demonstrated to be a powerful method for various γ-ray source classification tasks [14,20]. The novel aspect of our work is the use of BNNs, which allow us to quantify the uncertainty of the classification prediction.…”

The Third Fermi Large Area Telescope Catalog of Gamma-Ray Pulsars

“…Early successes in selecting pulsar candidates from among the unidentified LAT catalog sources exploited SED curvature versus flux variability correlations (Ackermann et al 2012b). Subsequent works used machine-learning methods to refine the selections (Lee et al 2012;Mirabal et al 2012;Saz Parkinson et al 2016;Wu et al 2018;Luo et al 2020;Finke et al 2021), or visual inspection and ranking of the LAT source spectra (Camilo et al 2015). Regardless of the ranking scheme, these lists of pulsar-like unassociated LAT sources have provided a large number of candidate pulsar positions that have been targeted by radio, X-ray, and gamma-ray searches.…”

A New Mid-Infrared and X-ray Machine Learning Algorithm to Discover Compton-thick AGN

“…We note that neural networks are another commonly used machine learning algorithm (see, e.g.,Finke et al 2021;Chainakun et al 2022; Zubovas et al 2022). We applied this technique to our data set and after finding the optimal configuration, yielded very similar results to those generated by our linear regression model.…”