[Clas12_software] Data and MC processed with AI-assisted tracking
Harut Avagyan
avakian at jlab.org
Tue Aug 24 09:46:04 EDT 2021
Dear All,
I put a few slides comparing the versions B and A, using the tracks
reconstructed and identified by EB, which are indeed real tracks
existing in the generated MC.
You can see from
https://userweb.jlab.org/~avakian/tmp/MC-comparison-B-vs-A.pdf
<https://userweb.jlab.org/~avakian/tmp/MC-comparison-B-vs-A.pdf>
that B has few percent more for e- and e-pion pairs for no background
version, but most importantly the number of real tracks reconstructed
with background, increases to ~20% for electron + 1 hadron. With
tighter and tighter fiducial cuts, that fraction seems to decrease. We
can check the impact of edges. In any case those are real tracks, with
possibly slightly worse energy and angular spreads.
Best,
Harut
On Mon, Aug 23, 2021 at 1:07 PM Raffaella De Vita
<Raffaella.Devita at ge.infn.it <mailto:Raffaella.Devita at ge.infn.it>> wrote:
Dear All,
Thanks a lot to Daniel and Stepan for analyzing the newly processed
data. If you are available, we could discuss these results (as well
as others that may be available) at this week software meeting.
Meanwhile, we have been doing some checks to understand the
difference Stepan found when comparing the J/psi yield to pass1 and
we can discuss that as well.
For anyone interested in looking at these data, we have also run
trains on the A and B samples using the last train configuration
adopted by RG-A. Outputs are available in
/volatile/clas12/users/clas12/mlt11/dst/trainA
/volatile/clas12/users/clas12/mlt11/dst/trainB
See /volatile/clas12/users/clas12/mlt11/dst/trainA/dstai/train/README.json
for the train configuration.
Best regards,
Raffaella
> On 23 Aug 2021, at 01:48, Valery Kubarovsky <vpk at jlab.org
> <mailto:vpk at jlab.org>> wrote:
>
> Dan,
> I don't see any fits in your plot. Do you count just the number of
> the events to compare A and B?
> Thanks,
> Valery
>
>
> ------------------------------------------------------------------------
> *From:*Clas12_software <clas12_software-bounces at jlab.org
> <mailto:clas12_software-bounces at jlab.org>> on behalf of Nathan
> Baltzell <baltzell at jlab.org <mailto:baltzell at jlab.org>>
> *Sent:*Saturday, August 21, 2021 16:42
> *To:*Daniel Carman <carman at jlab.org <mailto:carman at jlab.org>>
> *Cc:*clas12 software <clas12_software at jlab.org
> <mailto:clas12_software at jlab.org>>; Raffaella De Vita
> <raffaella.devita at ge.infn.it <mailto:raffaella.devita at ge.infn.it>>
> *Subject:*Re: [Clas12_software] [EXTERNAL] Data and MC processed
> with AI-assisted tracking
> Hello Daniel,
>
> Please use the A/B nomenclature provided and do not arbitrarily
> assign AI/CONV.
>
> -Nathan
>
>
>> On Aug 21, 2021, at 3:18 PM, Daniel Carman <carman at jlab.org
>> <mailto:carman at jlab.org>> wrote:
>>
>> Folks,
>>
>> I have compared the new cooking of RG-A fall 2018 inbending data
>> with conventional tracking and AI-assisted
>> tracking. The results of my study are attached in terms of 4
>> plots for MM(e’K+) using the skim14 kaon skim
>> files with the electron in the ECAL.
>>
>> UL: e’K+(FD) - AI has 19% more events
>> UR: e’K+(FD)p(FD) - AI has 28% more events
>> LL: e’K+(FD)p(CD) - AI has 14% more events
>> LR: e’K+(CD)p(FD) - AI has 18% more events
>>
>> Here FD implies hadron detection in the Forward Detector and CD
>> implies hadron detection in the Central Detector.
>> I have used the same run numbers in both studies. I also compared
>> same distributions from the recently cooked files
>> to the same runs with the original pass-1 cooking for the
>> convention tracking. The number of events for the different
>> distributions agrees to within a few percent.
>>
>> Regards,
>> Daniel
>>
>> *****************************************************************************
>> *
>> * Dr. Daniel Carman e-mail: carman at jlab.org <mailto:carman at jlab.org>
>> * Staff Scientist office: (757)-269-5586
>> * Jefferson Laboratory web: http://userweb.jlab.org/~carman
>> <http://userweb.jlab.org/~carman>
>> *
>> *****************************************************************************
>>
>>
>>
>>
>>
>>> On Aug 20, 2021, at 5:43 PM, Stepan Stepanyan <stepanya at jlab.org
>>> <mailto:stepanya at jlab.org>> wrote:
>>>
>>> Dear all,
>>>
>>> My studies of J/psi yield with new processed 31 runs are not
>>> complete, I am missing one run, 5126, for the skim1 of the
>>> trainA (reconA) but want to share the results that I have.
>>> Attached are 3 distributions of the e+e- invariant mass. The old
>>> one is what we have from the pass1 processing, using just the 31
>>> runs that have been processed for the AI-assisted tracking test.
>>> The new A is from the trainA (reconA) and the new B is from
>>> trainB (reconB). So, B has ~25% more J/psi’s than the pass1, A
>>> has about 25% less but, of course, with 1 run missing.
>>>
>>> Regards, Stepan
>>>
>>>
>>>
>>>
>>> > On Aug 20, 2021, at 7:11 AM, Raffaella De Vita
>>> <raffaella.devita at ge.infn.it
>>> <mailto:raffaella.devita at ge.infn.it>> wrote:
>>> >
>>> > Dear All,
>>> > Following the discussion at the software meeting of few weeks
>>> ago, significant statistics for both real data and MC (sidis)
>>> was processed with AI-assisted tracking to allow analyses of
>>> different final states and comparison to conventional tracking.
>>> >
>>> > For both real data and MC, two sets of DSTs files were
>>> produced, one based on conventional tracking and one based on
>>> AI-assisted tracking.
>>> >
>>> > The real data samples from RG-A Fall18 unbending are available in:
>>> > /volatile/clas12/users/clas12/mlt11/dst/reconA/
>>> > /volatile/clas12/users/clas12/mlt11/dst/reconB/
>>> >
>>> > The MC data samples, in RG-A Fall18 unbending configuration,
>>> are available in:
>>> > /volatile/clas12/users/devita/mlt11/reconA
>>> > /volatile/clas12/users/devita/mlt11/reconB
>>> > In these two folders, the nobg and bg folders contain
>>> reconstruction outputs with pure sidis events and after merging
>>> with 45nA background.
>>> >
>>> > Let me know if you have any question.
>>> > Best regards,
>>> > Raffaella
>>> >
>>> > _______________________________________________
>>> > Clas12_software mailing list
>>> >Clas12_software at jlab.org <mailto:Clas12_software at jlab.org>
>>> >https://mailman.jlab.org/mailman/listinfo/clas12_software
>>> <https://mailman.jlab.org/mailman/listinfo/clas12_software>
>>>
>>> <in_old31_557.pdf><in_newa30_557.pdf><in_newb31_557.pdf>_______________________________________________
>>> Clas12_software mailing list
>>> Clas12_software at jlab.org <mailto:Clas12_software at jlab.org>
>>> https://mailman.jlab.org/mailman/listinfo/clas12_software
>>> <https://mailman.jlab.org/mailman/listinfo/clas12_software>
>> <ai-conv-study.pdf><smime.p7s>
_______________________________________________
Clas12_software mailing list
Clas12_software at jlab.org <mailto:Clas12_software at jlab.org>
https://mailman.jlab.org/mailman/listinfo/clas12_software
<https://mailman.jlab.org/mailman/listinfo/clas12_software>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.jlab.org/pipermail/clas12_software/attachments/20210824/1d6b9256/attachment-0001.html>
More information about the Clas12_software
mailing list