[Rgc_analysis] [Clas12_calcom] first draft of pass1 review slides

[Gagik Gavalian]

Hi Silvia,

That’s fine, for the purpose it was made.
May be these points can be made more clear on the slides, so people are not confused.

But I do think it’s odd showing that MC reconstructs more tracks than generated, when background is not added.

Get Outlook for iOS<https://aka.ms/o0ukef>
________________________________
From: silvia <silvia at jlab.org>
Sent: Sunday, February 16, 2025 2:27:53 PM
To: Gagik Gavalian <gavalian at jlab.org>
Cc: Clas12 Calcom <clas12_calcom at jlab.org>; Rgc Analysis <rgc_analysis at jlab.org>; Rgc <rgc at jlab.org>
Subject: Re: [Clas12_calcom] first draft of pass1 review slides

Hi Gagik,
first of all, the purpose of this set of slides is to summarize the
conclusions to which we got (Raffaella, you, and I) when we decided
which network was best for RGC Spring23. So one has to focus on the
comparison Spring network vs summer network.
This said, if I recall correctly the numbers in slide 50 compare
with/without AI assisted tracking while the one in slide 53 compares
with/without denoising.
For the efficiencies above 1, I think it comes from the way I ran the
aiTracking code, I recall that there is an option to set the "zero"
point exactly at 1 but I didn't use it to make that plot.
Best regards,
Silvia


On 2025-02-14 12:43, Gagik Gavalian wrote:
> Hi Silvia,
>
> This may be an irrelevant question, but something strikes me as odd in
> the slides 50 and 53,
> According to slide 50, the gain in single particle reconstruction
> efficiency (for both charges) is about ~12%
>
> However, on slide 53, at 4 nA, it only shows 2-4% gain in efficiency
> (as far as one can tell from the graph).
> And extrapolated efficiency for 0 nA is above one for the data, which
> I guess normalization was done on AI, so
> it’s the normal procedure.
> However, what’s confusing is that on the MC plots, the efficiency is
> also above one, and it’s not calculated using
> extrapolation, but there is an actual point there. Do we reconstruct
> more tracks in MC when no background is added?
>
> — Gagik Gavalian
> — Staff Scientist
> — Jefferson Lab
> — email: gavalian at jlab.org
>
>> On Feb 14, 2025, at 12:06 PM, silvia via Clas12_calcom
>> <clas12_calcom at jlab.org> wrote:
>>
>> Dear Calcom and RGC,
>> I have prepared a first draft of the slides for the pass1 review:
>> https://box.in2p3.fr/s/BQHfiJnXY6xKmrA
>> If you could please check them, it would be great.
>> Things that will be added shortly:
>> - SIDIS analysis on the test runs
>> - Estimates of computing resources
>> - Back-up slides
>> - whatever has written "to be done" on the slide :-)
>> Thank you all for your contribution!
>> Best regards,
>> Silvia
>> _______________________________________________
>> Clas12_calcom mailing list
>> Clas12_calcom at jlab.org
>> https://mailman.jlab.org/mailman/listinfo/clas12_calcom
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.jlab.org/pipermail/rgc_analysis/attachments/20250216/18abb209/attachment.html>