Gravitational Wave Searches for High Mass Binaries in a Single Interferometer

Detecting gravitational waves (GWs) with a single detector, such as the LIGO-Virgo-KAGRA (LVK) detectors, can be challenging due to (near-)concurrences between transient noise artifacts (glitches) and real signals. Robust ways of identifying signals in such scenarios improves the detection rates and allows for a more accurate survey of source demographics. This project presents a ResNet based machine learning model aimed to identify GW events in a single detector even in the presence of glitches. We trained on signals with total masses of 50-120 solar masses, mass ratios from 1-10, and SNRs greater than 7. The ResNet model achieved an area under the Receiver Operating Characteristic curve (AUC) of 0.93, outperforming matched filtering with traditional and sine-Gaussian chi-square discriminator down-weighting (AUC = 0.90). The model consistently outperformed matched filtering in detecting signals, especially in concurrent cases, and is being integrated into LVK’s high-mass investigations for performing single-detector searches.