Binary Neutron Star Gravitational Wave Signal Classification using Machine Learning

The PyGRB offline analysis pipeline utilizes data from both Advanced LIGO and Advanced Virgo detectors in an effort to detect gravitational waves coincident with gamma ray bursts (GRBs). PyGRB offline currently utilizes a heuristic method called BestNR to distinguish between true gravitational wave signals and transient bursts of noise. This project explores an alternative method to classify gravitational wave signals using a neural network trained with data from chi-squared tests of each signal in offline GRB boxes. Initial testing using data from GRBs in the second half of the third observing run of Advanced LIGO and Advanced Virgo shows promising results, including an increased 90% confidence range of detection as well as a larger number of correctly classified injections compared to BestNR. Testing of different configurations will continue, including changing the GRBs used to train the model, utilizing subsets of the available chi-squared tests, as well as attempting to optimize the hyperparameters of the neural network in an effort to create the best performing model. We will present the status of these tests.