PARMIGGIANI, Nicolo'Nicolo'PARMIGGIANIBULGARELLI, ANDREAANDREABULGARELLIURSI, ALESSANDROALESSANDROURSIMacaluso, A.A.MacalusoDi Piano, A.A.Di PianoFIORETTI, ValentinaValentinaFIORETTIAboudan, A.A.AboudanBaroncelli, L.L.BaroncelliAddis, A.A.AddisTAVANI, MarcoMarcoTAVANIPITTORI, CarlottaCarlottaPITTORI2023-04-192023-04-1920230004-637Xhttp://hdl.handle.net/20.500.12386/34089AGILE 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 react to external science alerts received by other facilities. The AGILE anti-coincidence system (ACS) comprises five panels surrounding the AGILE detectors to reject background-charged particles. It can also detect hard X-ray photons in the energy range 50 - 200 keV. The ACS data acquisition produces a time series for each panel. The time series are merged into a single multivariate time series (MTS). We present a new Deep Learning model for the detection of GRBs in the ACS data using an anomaly detection technique. The model is implemented with a Convolutional Neural Network autoencoder (CNN) architecture trained in an unsupervised manner, 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. We calculated the associated p-value distribution, using more than $10^7$ background-only MTSs, to define the statistical significance of the detections. We evaluate the trained model with a list of GRBs reported by the GRBWeb catalog. The results confirm the model's capabilities to detect GRBs in the ACS data. We will implement this method in the AGILE real-time analysis pipeline.STAMPAenArticle10.3847/1538-4357/acba0ahttps://iopscience.iop.org/article/10.3847/1538-4357/acba0aFIS/05 - ASTRONOMIA E ASTROFISICA