Left: comparison of the AGN dN/dF^E-distributions described above. The blue histogram displays the actual distribution of 4FGL AGN-like sources whereas the orange histrogramme reflects the FL extrapolation to train the network, i.e. using a the threshold . The distribution with gray crosses shows the FL extrapolation to the lower threshold , which we only use as a verification data set. The gray pentagons denote the distribution arising from the PL extrapolation. Right: latitude distribution of PSR sources in 4FGL (blue) and an example mock catalog of 4FGL-like PSRs (orange). We model the latitude distribution of PSRs with a double Gaussian whose parameters are given by (A₁,μ₁,σ₁) = (0.11,0°,1.39°) for the first Gaussian and (A₂,μ₂,σ₂) = (0.012,0°,19.2°) for the second Gaussian. The quantities A_1/2 are the normalization constant of the two Gaussian distributions.