Please read these instructions before posting any event on Fermilab Indico

Indico search will be reestablished in the next version upgrade of the software: https://getindico.io/roadmap/


   

This search is only for public events. Restricted events are not available.

December 4, 2020
US/Central timezone

Graph Neural Networks for Reconstruction in DUNE

Dec 4, 2020, 12:00 PM
30m

Speaker

Jeremy Hewes (University of Cincinnati)

Description

The Deep Underground Neutrino Experiment is a next-generation long-baseline neutrino oscillation experiment, designed to measure current unknowns in neutrino oscillation phenomenology. Its high-resolution detectors benefit from ML-based reconstruction techniques, which currently provide the backbone of its projected sensitivity to primary physics goals. This talk will discuss the development of graph neural network (GNN) based low-level reconstruction techniques, using an attention message-passing network to classify the relationships between detector hits into different particle types. The current benchmark achieves 84% accuracy in graph edge classification, with more advanced techniques currently in development.

Presentation Materials