Search for glitches in gamma-ray pulsars with deep learning

Pulsar glitches are generally assumed to be an apparent manifestation of the superfluid interior of neutron stars. Most of them have been discovered and extensively studied by continuous monitoring of radio emission. The Fermi-LAT space telescope has revolutionized the field by uncovering a large population of gamma-ray pulsars. In this paper we employ the observations of gamma-ray pulsars to search for new glitches. We developed a method capable of detecting step-like frequency changes associated with glitches in sparse gamma-ray data. The method is based on the calculation of the weighted H-test statistics and consequent glitch identification by a convolutional neural network. The method demonstrates the high accuracy of the Monte Carlo set and is applicable to searching for pulsar glitches in real gamma-ray data