Flexible and fast estimation of binary merger population distributions with an adaptive kernel density estimator

The LIGO Scientific, Virgo and KAGRA Collaborations recently released the third gravitational wave transient catalog or GWTC-3, significantly expanding number of signals. To address the---still uncertain---formation channels source compact binaries, their population properties must be characterized. computational cost Bayesian hierarchical methods employed thus far scales with size event catalogs, such have until assumed fixed functional forms for distribution. Here we propose a fast flexible method to reconstruct LIGO-Virgo merging black hole (BH) binaries without assumptions. For sufficiently high statistics low individual measurement error (relative scale features) kernel density estimator (KDE) reconstruction distribution will accurate. We improve accuracy flexibility KDE finite using an adaptive bandwidth (awKDE). apply awKDE publicly parameter estimates 44 significant (69) BH binary mergers in GWTC-2 (GWTC-3), combination polynomial fit search sensitivity, obtain nonparametric estimate mass distribution, compare methods. also demonstrate robust peak detection algorithm based on use it calculate significance apparent around $35\text{ }\text{ }{\mathrm{M}}_{\ensuremath{\bigodot}}$. find is very unlikely occurred if true featureless power-law (significance $3.6\ensuremath{\sigma}$ confident BBH events, $3.0\ensuremath{\sigma}$ GWTC-3 events).