Zoom coordinates: note the new link!!!
ePIC Reconstruction meeting, May 21
Host: Derek
Note taking: Shujie
-------------------------------------
1. AOB:
a. Announce Simon Gardner as the user learning WG liaison
b. primary vertex recon status
c. EICUG workfest ideas:
i. workfest of documentation.
John: this may be relevant for only a small group of people, maybe fit better with some S&C meeting.
Markus: get people in one room, and learn from users on missing docs and what is desired. Collaboration meeting is a unique opportunity for people to travel.
ii. general workfest session on reconstruction status, and also discuss the plan on PID
------------
2. FF ZDC (HCal) reconstruction status (Sebouh)
Two approaches:
a. topological clustering:
existing imaging calo algorithm. multiple hits along z, need to optimize the clustering parameters
b. Graph Neural Network
better resolution, less bias, can be used for multi-neutron clustering.
-------------
Discussion:
Miguel: would be nice to adding noise to ZDC
Derek: joint meeting b/w Kolja, Miguel, Derek on noise implementation
Markus: detailed noise model is not priority for TDR
Dimitrii: how GNN is set up, and its memory consumption?
Miguel: still need to develop a cpp code to handle the EICrecon fit to GNN input transition
Alex: use pytorch for model, Sakib is working on another GNN plugin and also code to link EICrecon to trained model, which may be useful here as well.
Derek: combine EMCal and HCal
Miguel: try to reconstruction neutron from labmda with HCal only, once it's under contral, then add ECal
Alex: you need imaging capability for lambda recon without vertexing info. EMCal may not be that helpful, and may smear the showers a little bit which can be a simple study to do.
3. efinder update
Derek (Tristan can not attend):
efinder will move to using real (instead of truth) info for July campaign