Classification of Fermi-LAT blazars with Bayesian neural networks

The use of Bayesian neural networks is a novel approach for the classification gamma-ray sources. We focus on Fermi-LAT blazar candidates, which can be divided into BL Lacertae objects and Flat Spectrum Radio Quasars. In contrast to conventional dense networks, provide reliable estimate uncertainty network predictions. explore correspondence between effect data augmentation. find that robust classifier with estimates are particularly well suited problems based comparatively small imbalanced sets. results our candidate valuable input population studies aimed at constraining luminosity function guide future observational campaigns.