Information on precautionary site access restrictions to Fermilab's Batavia site in response to coronavirus/COVID-19
We continue to review all events currently planned for the next sixty days and organizers will be notified if their event must be canceled, postponed, or held remotely. Please, check back on Indico during this time for updates regarding your meeting specifics.
As DOE O 142.3A, Unclassified Foreign Visits and Assignments Program (FVA) applies not only to physical access to DOE sites, technologies, and equipment, but also information, all remote events hosted by Fermilab must comply with FVA requirements. This includes participant registration and agenda review. Please contact Melissa Ormond, FVA Manager, with any questions.

indico search will be reestablished in the next version upgrade of the software: https://getindico.io/roadmap/
For public events you may use either https://library.fnal.gov/indico-search/ or your browser's search engine: "your search string" site:indico.fnal.gov

18-19 June 2018
Fermilab, Wilson Hall
US/Central timezone

Accelerating Feldman-Cousins on NOvA using NERSC Supercomputers

Jun 19, 2018, 5:30 PM
15m
One West (Fermilab, Wilson Hall)

One West

Fermilab, Wilson Hall

Oral Presentation Long Baseline Neutrino Program

Speaker

Derek Doyle

Description

When fitting to data with low stats and near physical boundaries, extra measures need to be taken to ensure proper statistical coverage. The method NOvA uses is called the Feldman-Cousins procedure, which entails fitting thousands of independent pseudoexperiments to generate acceptance intervals that are then used to correct our fits. The scale required by the Feldman Cousins procedure makes it extremely computationally intensive. In past analyses, it has taken up to several weeks to complete, bottlenecking our final results. Here I present recent work by members of the NOvA experiment and the SciDAC4 collaboration to enable the use of the super-computing facilities at NERSC to process our Feldman-Cousins corrections over 150x faster, allowing us to do more studies, increase the precision of our fits, and produce results quickly.

Primary author

Presentation Materials