Insights into the predictions of a machine learning classifier for gravitational-wave events

Nayyer Raza ( Université McGill )

Multi-messenger observations of both the gravitational waves and electromagnetic emission from compact object mergers can give unique insights into the structure of neutron stars, the formation of heavy elements, and the expansion rate of the universe. With the LIGO-Virgo-KAGRA (LVK) gravitational-wave detectors having recently begun the second half of their fourth observing run (O4b), it is an exciting time for detecting such events. However, assessing whether to follow up a candidate gravitational-wave event given limited telescope time and resources is challenging; the candidate can be a false alert due to detector glitches, or not have any detectable electromagnetic counterpart even if it is real. GWSkyNet-Multi is a machine learning model developed to facilitate follow-up decisions by providing real-time classification of candidate events, using limited information released in LVK rapid public alerts. While the model is successful, as a deep learning neural network its inner workings are not transparent and thus its classifications can be hard to interpret. In a recent study we take a deep dive into explaining GWSkyNet-Multi, using systematic model input perturbations to identify the information that is the strongest predictor for its output. We leverage these findings to understand events misclassified by GWSkyNet-Multi in LVK’s third observing run, and inform potential avenues for further optimizing the model for O4b events.