Speaker
Guanqun Ge
(Columbia University)
Description
The Deep Underground Neutrino Experiment (DUNE) is a planned, next-generation experiment that will use the liquid-argon time projection chamber (TPC) technology to study three-neutrino oscillations and search for rare physics processes. DUNE will have four far detector modules, each 10 ktons in fiducial mass. From just one of those modules, the TPC raw data will be streamed out of the DUNE detector at a rate faster than 9 terabits per second. The raw data from the detector can be visualized in the form of high-resolution (11.5 megapixels) images, streamed at a frame rate of 67,000 per second. This invites the application of deep neural networks for online or real-time image classification as a trigger method. One of the main physics goals of DUNE is the study of supernova (SN) neutrinos from a galactic or nearby supernova burst, shall that happen during the experiment's lifetime. The rareness of such burst requires an efficient trigger system in DUNE to differentiate a potential SN neutrino event from other backgrounds or noise in the vast data stream. This talk will present ongoing efforts to demonstrate the feasibility of a machine learning-based trigger at DUNE.
Primary author
Guanqun Ge
(Columbia University)
