Extracting the Galactic Center excess’ source-count distribution with neural nets

The two leading hypotheses for the Galactic Center Excess (GCE) in $\textit{Fermi}$ data are an unresolved population of faint millisecond pulsars (MSPs) and dark-matter (DM) annihilation. dichotomy between these explanations is typically reflected by modeling them as separate emission components. However, point-sources (PSs) such MSPs become statistically degenerate with smooth Poisson ultra-faint limit (formally where each source expected to contribute much less than one photon on average), ambiguity that can render questions whether PS-like or Poissonian nature ill-defined. We present a conceptually new approach describes PS unified manner only afterwards derives constraints component from so obtained results. For implementation this approach, we leverage deep learning techniques, centered around neural network-based method histogram regression expresses uncertainties terms quantiles. demonstrate our robust against number systematics have plagued previous approaches, particular DM / misattribution. In data, find GCE described median source-count distribution (SCD) peaked at flux $\sim4 \times 10^{-11} \ \text{counts} \text{cm}^{-2} \text{s}^{-1}$ (corresponding $\sim3 - 4$ counts per PS), which would require $N \sim \mathcal{O}(10^4)$ sources explain entire excess (median value = \text{29,300}$ across sky). Although faint, SCD allows us derive constraint $\eta_P \leq 66\%$ fraction $\eta_P$ 95% confidence, suggesting substantial amount due PSs.