Orateur
Description
Multiwavelength and multimessenger observations, combining data across the electromagnetic spectrum with high-energy neutrinos, provide a uniquely powerful probe of the physical processes at work in the relativistic jets of blazars. Recent campaigns and coincident neutrino detections have reinforced the role of blazars as prime laboratories for studying particle acceleration and radiation in extreme environments, while simultaneously highlighting the need for modeling frameworks that can consistently interpret these rich and heterogeneous datasets. However, physically motivated leptonic, hadronic, and hybrid radiative models remain computationally demanding, making comprehensive parameter-space exploration and statistically robust fitting of multiwavelength and multimessenger data prohibitively expensive. In this presentation, I introduce a novel approach based on convolutional neural networks (CNNs) which accelerates broadband blazar emission modeling. The CNN is trained on synthetic spectra produced by the SOPRANO numerical code and accurately reproduces the radiative output from electrons, protons, and secondary particles across the full electromagnetic range, while preserving the physical structure of the underlying model. This approach effectively transforms computationally intensive radiative calculations into a fast surrogate model suitable for large-scale parameter scans and rigorous statistical inference. I demonstrate the performance and scientific impact of this method through joint modeling of multiwavelength and multimessenger spectral energy distributions of several blazars, showing that it enables efficient and self-consistent constraints on the physical conditions in the jet and a coherent interpretation of both photon and neutrino emission.
