Assessment of normalizing flows for parameter estimation on time-frequency representations of gravitational-wave data

Abstract: The speed-up of parameter estimation is an active field of research in gravitational-wave data analysis. In this paper we present GP12, a deep-learning method that merges residual networks and normalising flows into a general-purpose, image-based estimator of binary black hole (BBH) parameters. Building on our early work we map BBH spectrograms from the Advanced LIGO and Advanced Virgo detectors to colour channels in an RGB image amenable to be processed with residual networks. GP12 is trained on simulated data for BBH mergers obtained with the $\texttt{IMRPhenomPXHM}$ waveform approximant and tested for all three-detector events from the GWTC-3 and GWTC-2.1 catalogs reported by the LIGO-Virgo-KAGRA (LVK) collaboration. Overall, our model yields good agreement with the LVK results over most parameters, with the worst performances found in the estimation of the luminosity distance and of the chirp mass. Our simple and fast-trainable model can produce large amounts of posterior samples in a few seconds, complementing existing approaches with normalising flows based on time or frequency representation of gravitational-wave data. We also discuss current shortcomings of our model and possible improvements for future extensions (e.g. including noise conditioning from the detectors' PSD or augmenting the number of trainable parameters to enhance expressivity).