Testing the robustness of simulation-based gravitational-wave population inference

Gravitational-wave population studies have become more important in gravitational-wave astronomy because of the rapid growth observed catalog. In recent studies, emulators based on different machine learning techniques are used to emulate outcomes synthesis simulation with fast speed. this study, we benchmark performance two that learn truncated power-law phenomenological model by using Gaussian process regression and normalizing flows see which one is a capable likelihood emulator inference. We characteristic comparing their inference mock real observation data. Our results suggest can recover posterior distribution up 300 injections. The also underestimates uncertainty for some distributions On other hand, has poor same task only be effectively low-dimension cases.