21–26 Jun 2026
University of California, Irvine
US/Pacific timezone

AI/ML-based Real-time Anomaly Detection in DUNE for Supernova Burst Neutrinos

Not scheduled
20m
Conference Center (University of California, Irvine)

Conference Center

University of California, Irvine

Poster Supernova Neutrinos Poster session 2

Speaker

Seon-Hee (Sunny) Seo (Fermilab)

Description

The Deep Underground Neutrino Experiment (DUNE) is currently under construction with far detectors consisting of 4 liquid argon time projection chamber (LArTPC) modules at SURF (South Dakota Underground Research Facility) and a near detector complex with neutrino beam production at Fermilab to unambiguously determine neutrino mass ordering, to discover and precisely measure Charge-Parity (CP) violation phase in leptonic sector, to search for Beyond Stand Model (BSM) physics, and to study solar and supernova burst neutrinos.

Anomalies in this project are classified in three categories: new physics signals, supernova burst neutrinos, and detector malfunction. We report here on promising early studies toward an Artificial Intelligence/Machine Learning-based real-time anomaly detection system, using a prototype autoencoder model currently under development. Additionally, the current status of an improved model and its performance will be presented. The model will be evaluated not only for its sensitivity to supernova neutrinos, but also to BSM physics signals and detector malfunctions. We will also consider how such a real-time algorithm might be used in DUNE.

Authors

Anselm de Jonge (KIP Heidelberg) Dr Loredana Gastaldo (Kirchhoff Institute for Physics, Heidelberg University) Seon-Hee (Sunny) Seo (Fermilab)

Presentation materials