Fast neural-net based fake track rejection

Goal

Build a new fast neural-net based fake track rejections, which is fast enough to run in the trigger. And thereby reduce the workload for PID and combinatorics and thus save time in the total HLT.

Proponents

Michel De Cian, Stephen Farry, Paul Seyfert, Sascha Stahl

Previous results

LHCb-INT-2012-025

Review

comments in websvn

• 26.08.2016: The first draft of the note sent to the T&A convenors by Paul.
• 21.09.2016: 1st set of comments from Francesco Polci
• 27.10.2016: 1st set of comments from Michel De Cian.
• 25.01.2017: 2nd draft of the note sent to the T&A convenors by Paul with partial reply to previous comments
• 22.02.2017: an updated (2.1) draft sent by Paul
• 04.08.2017: an updated (2.2) draft sent by Paul
• 09.08.2017: 1st set of comments from Agnieszka Dziurda
• 14.08.2017: 3rd draft sent by Paul with Agnieszka's suggestions.
• 05.09.2017: updated (3.1) draft with ref [3] filled in
• 20.10.2017: updated (3.2) draft with references to internal documents removed
• 12.12.2017: Tuesday meeting presentation

Documentation

weblink: websvn, svn: svn+ssh://svn.cern.ch/reps/lhcbdocs/Users/pseyfert/GP

• The documentation can be found here:
• Verison 1.0: LHCb-PUB-2017-011-v1.0
• Verison 2.0: LHCb-PUB-2017-011-v2.0
• Version 2.1: LHCb-PUB-2017-011-v2.1
• Version 2.2: LHCb-PUB-2017-011-v2.2
• Version 3.0: LHCb-PUB-2017-011-v3.0
• Version 3.1: LHCb-PUB-2017-011-v3.1
• Version 3.2: LHCb-PUB-2017-011-v3.2
• released version on CDS: https://cds.cern.ch/record/2255039

Software

Link to the package:

• training and development
• application in REC

-- AgnieszkaDziurda - 2017-08-28