Powered by Indico
Maximizing the number of detections in matched filter searches for compact binary coalescence gravitational wave signals requires a model of the source population distribution. In previous searches using the PyCBC framework, sensitivity to the population of binary black hole mergers was improved by restricting the range of filter template mass ratios and using a simple one-dimensional population model. However, this approach does not make use of our full knowledge of the population and cannot be extended to a full parameter space search. Here, we introduce a new ranking method, based on kernel density estimation with adaptive bandwidth, to accurately model the probability distributions of binary source parameters over a template bank, both for signals and for noise events. We demonstrate this ranking method by conducting a search over LIGO-Virgo O3 data for BBH with unrestricted mass ratio, using a signal model derived from previous significant detected events. We achieved over 10% increase in sensitive volume for a simple power-law simulated signal population, compared to the previous BBH search. Correspondingly, with the new ranking, 8 additional candidate events above an inverse false alarm rate threshold of 0.5 yr are identified.