Blazar classification from multi-wavelength data using Deep Learning

TL;DR

This paper develops a deep learning approach to classify uncertain gamma-ray sources as blazar subclasses using multi-wavelength data, enhancing the understanding of the gamma-ray blazar population.

Contribution

It introduces a neural network model trained on selected multi-wavelength features to accurately classify blazar candidates of uncertain type.

Findings

High classification accuracy achieved for BCUs

Effective feature selection improves model performance

Method enhances catalog completeness

Abstract

The Fermi Large Area Telescope (Fermi-LAT) has detected more than 7,000 gamma-ray sources, a significant fraction of which are identified as blazars, while a comparable number remain classified as blazars of uncertain type (BCUs) or are unassociated with counterparts at other wavelengths. The absence of complete multi-wavelength spectral information presents a major obstacle to robust source classification, despite such data providing the most reliable means of understanding blazar properties. In this work, we focus on classifying BCUs into the two primary blazar subclasses, flat-spectrum radio quasars (FSRQs) and BL Lacertae objects (BL Lacs), using a feed-forward artificial neural network (ANN) trained on multi-wavelength observational parameters. We first identify the most informative features by quantifying their information content and then use these features to train the ANN, whose performance is evaluated using a k-fold cross-validation strategy to ensure robust generalization. The trained model is subsequently applied to classify BCUs into BL Lacs and FSRQs. Our results demonstrate that machine learning-based classification using a carefully selected set of multi-wavelength parameters offers an efficient and reliable approach for resolving the nature of BCUs and improving the completeness of the gamma-ray blazar population in Fermi-LAT catalogs.