Abstract:
GW observations are now revealing the population of black holes via the detection of binary black hole mergers, and the numbers are set to grow rapidly in the coming years: the distribution of astrophysical properties of binary black holes provides a key to interpret their formation channels. In this regard, scientists have proposed a variety of models for the black hole distribution, broadly classifiable into two categories: parametric models, informed by the underlying physics but bound to specific functional forms, and non-parametric methods, flexible but lacking immediate physical interpretation.
During this talk I will present a way to augment parametric models using a non-parametric method, namely a mixture model of an agnostic channel and one or more parametric models. This way, we can take the best of both worlds: the inference will be based on the understanding of physical processes encoded in the parametric models, whereas the non-parametric component will account for the observations that cannot be interpreted in terms of parametric channels. The inferred "missing part" of the distribution can therefore be used to guide the further development of physical models.
