Epic-backward-hcal-l mailing list

[Leszek Kosarzewski <leszek.kosarzewski AT gmail.com>]

Dear Colleagues

During today's meeting we had an update by David and discussion on the next steps to do a realistic study of neutral jet reconstruction. For details, please have a look at the plan, I sent via the calorimetry list. I also have good news. The group from UIUC wants to join and they have experience with calorimeters. This means that next week we can have a discussion on tasks and how to share the work.

For now the summary of today's discussion is:

ML study:

1. Strange drop of reconstructed energy at E~2 GeV - we need to investigate or generate more events. Maybe using log(E) as an input can help. It could be that the network tunes to the low energy showers and completely misses the high energy ones. I contacted Daniel-Han for help.

2. We need to simulate more events using a cluster. Here is my set of condor scripts and SLURM tutorial:

https://github.com/lkosarz/ePICcondorScripts

https://indico.bnl.gov/event/18374/#5-example-of-job-submission-sc

In principle ePIC containers can be used at any cluster, as they are easily shared.

Realistic neutral jet reconstruction and position resolution study:

1. Get track-cluster matching to work - I will discuss it with Daniel on Monday

2. We may need to look at full DIS events

3. Relate reconstructed and MC truth information to understand which particles produced neutral clusters

4. Truth clusters seem to use the first hit contribution to point to MC particle, the rest may be neglected

5. Please think how to do these studies and we can discuss more next week once we get feedback from calorimetry experts

Best regards, Leszek

czw., 9 maj 2024 o 14:49 Leszek Kosarzewski <leszek.kosarzewski AT gmail.com> napisał(a):

Dear Colleagues

Let's have a DSC meeting tomorrow at 1000h. During the last calorimetry meeting we had a strong presence and it was very good work, congratulations! I will send the responses to the calorimetry list soon (please look there) and we can plan next steps. Looks like the group wants us to include track-cluster matching which wasn't really available until recently. Let's discuss it during the meeting tomorrow:

1. Update from David on machine learning

2. Followup discussion from the calorimetry meeting

Best regards, Leszek

pt., 3 maj 2024 o 13:43 Leszek Kosarzewski <leszek.kosarzewski AT gmail.com> napisał(a):

FYI:

Here's the paper with the weight formula dependent on energy and segment size: https://iopscience.iop.org/article/10.1088/1748-0221/11/07/P07001

Best regards, Leszek

pt., 3 maj 2024 o 13:40 Leszek Kosarzewski <leszek.kosarzewski AT gmail.com> napisał(a):

Dear Colleagues

During the DSC meeting today, we had discussions on the topics listed below. It looks like at the upcoming calorimetry meeting on Wednesday, we are going to have 3 presentations in total. This is very good and underlines the progress made by each of you. We are definitely on a right track.

1. David showed first results of ML with HCal only. The plan now is to generate particles in the backward direction only and use fixed energy ranging from 0.3-several GeV.

2. Subhadip briefly discussed changes to his presentation. We noticed there is a rotation of geometry and it also rotates the x,y segmentation. This is caused by the following code: https://github.com/eic/epic/blob/ed8f4cf3d4119d0488c77fb9484931cd33b018c9/src/PolyhedraEndcapCalorimeter2_geo.cpp#L122

I will update that.

Also the position resolution depends on the weight in the clustering algorithm. This weight in turn depends on energy and geometry. There is a formula describing the dependence, which we can use for guidance and set it for neutron kinetic energy E=1.44 GeV (2.38 GeV total energy). This is the mean energy of neutrons in backward HCal coming from DIS events+secondaries. Could you add a link to the paper containing the formula? I couldn't find it in Awes et al.

3. Alexandr showed his slides on the position resolutions study, which are ready for the calorimetry meeting. The position resolution looks constant vs. radial distance from the beam, so he picked up a single point to shoot particles. In this way he obtained the scaling of resolution vs. energy and tile size. The resolution follows dR_MPV=a/\sqrt{E}+b, where dR_MPV comes from Landau fit. The RMS behaves similarly and also gives similar values up to 25 cm x 25 cm tiles.

4. Brian suggested that we make sure track projections and cluster matching works first to do a realistic identification of neutral clusters. Next we can use existing DIS events pure PYTHIA files as an input to npsim with updated geometry and later eicrecon. This can be done even at OSU cluster or ASC. I was able to find these files:

/work/eic2/EPIC/MC_input/SIDIS/pythia6/ep-18x275/

/work/eic2/EPIC/EVGEN/DIS/NC/

I'm not sure if these are correct, but please confirm.

Best regards, Leszek

czw., 2 maj 2024 o 14:50 Leszek Kosarzewski <leszek.kosarzewski AT gmail.com> napisał(a):

Dear Colleagues

We are going to have a DSC meeting on Friday at 1000h. Since we plan to present updates at the calorimetry meeting next week, this will be a good opportunity to review the presentations. However, I would like to start with regular updates.

1. David - progress update.

2. Alexandr - can you make it with the presentation for calorimetry meeting next week? If yes, tomorrow we can discuss it.

3. Subhadip - please send us the slides if you are ready.

4. Brian - Here we may need to start thinking about a few things:

 a) How the tile size may affect the performance of jet reconstruction with neutral hadrons

 b) On slide 11 of your presentation there seems to be a large jump of resolution at eta=-2.8. We need to understand where it comes from. It might be related to overlap with EMCal.

 c) If we use EMCal as an additional neutral veto, then Pi0 clusters may be an issue here. They would also show up as neutral.

Best regards, Leszek

niedz., 28 kwi 2024 o 20:56 Leszek Kosarzewski <leszek.kosarzewski AT gmail.com> napisał(a):

Dear Colleagues

During our meeting on Friday we discussed a few things:

1. Brain will give a presentation at the calorimetry meeting in 2 weeks to discuss it with experts.

2. David presented a first look at the shower reconstruction with ML. The results are surprisingly good. We need to check if the LFHCAL insert data might've been included by accident.

- simulate particles in nHCal acceptance only

3. Subhadip presented updated slides on clustering - we need to give an update at the calorimetry meeting too. Could you send me your slides from the meeting? We also need to see the updated slides to be presented at the calorimetry meeting. Subhadip, did you contact Oleg about your presentation?

4. The position resolution study also needs to be presented most likely this week. Alexandr, did you contact Oleg as well?

Best regards, Leszek

śr., 24 kwi 2024 o 23:02 Leszek Kosarzewski <leszek.kosarzewski AT gmail.com> napisał(a):

Dear Colleagues

We are going to have a backward HCal DSC meeting at 1000h BNL time on Friday. Please find the link below:

https://osu.zoom.us/j/99011057540?pwd=QmNhTDBrNmMyckdWYnJqcmFLMHl5QT09

The plan is to discuss presentations for the next week's calorimetry meeting. David will also give an update on the shower reconstruction with machine learning.

Brian, you are also welcome to join if you have an update. I will contact Oleg about your presentation for the calorimetry meeting next week.

Best regards, Leszek