Choose timezone
Your profile timezone:
FPD Seminar
Using Next-Generation Detectors With Graph Neural Networks
by
→
America/Los_Angeles
Zoom-only
Zoom-only
Description
Join from PC, Mac, Linux, iOS or Android: https://stanford.zoom.us/j/94642768923
Password: 134699
Or iPhone one-tap (US Toll): +18333021536,,94642768923# or +16507249799,,94642768923#
Or Telephone:
Dial: +1 650 724 9799 (US, Canada, Caribbean Toll) or +1 833 302 1536 (US, Canada, Caribbean Toll Free)
Meeting ID: 946 4276 8923
Password: 134699
International numbers available: https://stanford.zoom.us/u/adCshhWfqP
Meeting ID: 946 4276 8923
Password: 134699
SIP: 94642768923@zoomcrc.com
Password: 134699
As High Energy Physics makes progress towards ever high energies and smaller distance scales in its quest to describe the fundamental properties of nature, it requires ever-more precise detectors. In the efforts towards the HL-LHC and future colliders, detector instrumentation like trackers and calorimeters are using finer detector segmentation and precision timing information to deal with large numbers of overlapping collisions, and to improve final measurements. However, this comes with a steep computational cost when using traditional algorithms. One method of ameliorating these costs while maintaining or improving physics performance is to use a recent development in machine learning called Graph Neural Networks (GNNs). GNNs combine aspects of traditional programming with neural networks to create significantly more powerful neural networks that can also operate on large structured datasets representing detectors with irregular geometries, as often found in particle physics. In this talk I will discuss uses for the capabilities of these new detectors in the context of these exciting new algorithms and give demonstrations of current research in this direction.
