Searching for Long-Duration Transient Gravitational Waves: Convolutional Neural Networks Applied to Glitching Pulsars
摘要
Besides compact binary mergers and other sources, long-duration quasi-monochromatic signals from spinning deformed neutron stars have long been one of the prime targets of ground-based gravitational-wave detectors. Glitching pulsars in particular can be a source of such signals that are not quite as persistent as those from more quiescent neutron stars, but could be detectable on timescales of hours to months. Within the framework of the g2net COST action, at the University of the Balearic Islands a project has been pursued to develop a hybrid search approach for such signals that combines intermediate products from a matched-filter search over a bank of frequency-evolution templates with neural networks trained to identify signals of varying start time, duration and amplitude evolution. Here we summarise the motivation for this project and the results originally presented in Modafferi, Keitel & Tenorio 2023, Physical Review D 108, 023005 [1].