Self-Supervised ConvLSTM for Fermi Large Area Telescope Transient Detection

2026-05-21

AI Summary

This work addresses the challenge of automatically detecting gamma-ray transients—such as flares and gamma-ray bursts—in Fermi-LAT survey data, which is hindered by complex background modeling and ambiguous definitions of anomalous behavior. The study presents the first application of a self-supervised ConvLSTM architecture to decade-scale synthetic time-series sky maps, trained end-to-end on simulated daily all-sky count and exposure map sequences to learn normal spatiotemporal evolution patterns. Anomaly detection is achieved through pixel-level residual statistics, augmented by spatial consistency filtering to enable unsupervised transient identification. Applied to real Fermi-LAT data, the method successfully recovers localized excesses consistent with known highly variable sources or transient events, establishing a reproducible, unsupervised anomaly detection benchmark for long-term gamma-ray sky surveys.

Abstract

We present a framework for detecting transient gamma-ray phenomena in a controlled environment by combining end-to-end simulations of the Fermi-LAT sky with self-supervised spatio-temporal deep learning. We generate a ten-year synthetic Universe with gtobssim and process the simulated events into daily all-sky maps of counts and exposure, obtaining a time-ordered sequence that mirrors the structure of Fermi-LAT observations. To model the nominal evolution of the sky, we employ a Convolutional Long Short-Term Memory (ConvLSTM) network that operates directly on map sequences, preserving spatial locality while learning temporal dependencies. The model is trained to reconstruct expected emission, and departures from the learned baseline are quantified through pixel-wise mean-squared residual maps. We then define statistically motivated anomaly criteria by estimating per-pixel thresholds from the residual distribution on the training set, and we enforce spatial coherence via local filtering to suppress isolated fluctuations. The ConvLSTM is then deployed as trained predictor on Fermi-LAT daily maps, where the sky can depart from the nominal behavior because of genuine astrophysical variability and instrumental non-stationarities. The resulting pipeline flags localized, time-dependent excesses consistent with high-variable sources or transient events (e.g., flares or GRBs) and provides a benchmark for evaluating anomaly-detection strategies on long-duration, Fermi-LAT-like datasets.

Problem

• transient detection
• gamma-ray astronomy
• anomaly detection
• Fermi-LAT
• time-domain astrophysics

Innovation

• Self-supervised learning
• ConvLSTM
• Transient detection
• Anomaly detection
• Fermi-LAT