Preliminary Results of a Deep Learning Anomaly Detection Method to Identify Gamma-Ray Bursts in the AGILE Anticoincidence System

AGILE is a space mission launched in 2007 to study X-ray and gamma-ray astronomy. The AGILE team developed real-time analysis pipelines to detect transient phenomena such as Gamma-Ray Bursts (GRBs) and to react to external science alerts received by other facilities. The AGILE anti-coincidence system (ACS) comprises five panels (four lateral and one on the top) that surround the AGILE detectors to reject background charged particles. It can also detect hard X-ray photons in the energy range 50 - 200 KeV. The acquisition of the ACS data produces a time series for each panel. These time series can be merged in a single multivariate time series (MTS). We present in this work a new Deep Learning model for GRBs detection in the MTSs, generated by the ACS, using an anomaly detection technique. The model is implemented with a Deep Convolutional Neural Network autoencoder architecture. We trained the model with an unsupervised learning algorithm using a dataset of MTSs randomly extracted from the AGILE ACS data. The reconstruction error of the autoencoder is used as the anomaly score to classify the MTS. If the anomaly score is higher than a predefined threshold, the MTS is flagged as a GRB. The trained model is evaluated using a list of MTSs containing GRBs. The tests confirmed the model's ability to detect transient events, providing a new promising technique to identify GRBs in the ACS data that can be implemented in the AGILE real-time analysis pipeline.
Comment: 4 pages, 1 figure, proceedings of the ADASS XXXI (2021) conference, to appear in ASP Conference Serie