Real-time processing of new gamma-ray observations

Investigate the real-time processing of new very-high-energy gamma-ray observations using the convolutional variational autoencoder framework while retaining fine separation of faint signals mixed with stronger point sources.

Background

The paper introduces a convolutional variational autoencoder for separating signal and background in three-dimensional gamma-ray count data without assuming the signal multiplicity or morphology. The authors note that the framework could be extended to real-time processing of new observations and to the separation of faint signals from stronger point sources, but this functionality is not explored in the paper.

References

It is also extendable to real-time processing of new observations, whilst allowing for fine separation of faint signals mixed with stronger point sources., however exploration of this functionality is left for future work.

— Convolutional non-parametric Gamma-Ray Signal and Background Separation  (2609.30917 - Betterman et al., 25 Sep 2026) in Section 2.1, “Formalism and Method”