Searching for previously unknown γ -ray sources is a key objective of the Fermi LAT Collaboration. We address this challenge by clustering the directions of high-energy photon emissions detected by the Fermi spacecraft’s Large Area Telescope (LAT). Candidate sources are detected by analyzing the excess mass within discrete, high-density regions, allowing us to discriminate them from the diffuse γ -ray background that pervades the entire sky. Density estimation is performed nonparametrically using binned directional kernel methods applied to a sphere mesh. Source detection is facilitated by partitioning the problem into separate subregions of the sphere, delimited by empty bins, which results in a substantial gain in computational efficiency. • A scalable nonparametric method searches for gamma-ray sources in a diffuse background. • It leverages methodological properties to enhance computational efficiency. • The approach handles large-scale data distributed over the entire sphere. • It detects heterogeneous and unevenly represented signals without predefined models. • The method easily applies to further application settings and data problems.
Freni et al. (Sun,) studied this question.