[Hps-analysis] HPS Analysis meeting March 5 @ 9am/noon PST/EST

Wed Mar 6 16:25:22 EST 2019

Hi all,

An old idea - why we cannot use beam background from the data, using the 
random trigger events?

Stepan

On 3/6/19 3:27 PM, Solt, Matthew Reagan wrote:
>
> Hi Takashi,
>
>
> Thanks for thinking about this. You are correct in saying that a x10 
> sample of tritrig-wab-beam is computationally difficult (if not 
> impossible using our resources). However, I think training on a x10 
> sample of tritrig is sufficient. The goal of these ML studies is to 
> distinguish between multiple scattered tracks that produce a 
> downstream vertex (from a prompt trident) and a true displaced vertex. 
> So on that principle alone, I think a x10 sample of tritrig should be 
> enough for training.
>
>
> But of course we have to take wabs and beam backgrounds into account 
> somehow. In principle, I will find a way to get rid of the tracks that 
> pick up the wrong hit due to a beam background (a more sophisticated 
> isolation cut), and make it such that wabs are not such a big deal for 
> the vertexing. I will of course need to justify that these will not be 
> backgrounds in the vertexing analysis (with the ML method). One way to 
> do this is to test on the full 100% tritrig-wab-beam sample. I think 
> this should be enough to justify just training on a very large sample 
> of pure tridents, but someone may come with a counter argument.
>
>
> More ideas are welcome. Thanks.
>
>
> Matt Solt
>
> ------------------------------------------------------------------------
> *From:* Hps-analysis <hps-analysis-bounces at jlab.org> on behalf of 
> Maruyama, Takashi <tvm at slac.stanford.edu>
> *Sent:* Wednesday, March 6, 2019 11:54:47 AM
> *To:* hps-analysis at jlab.org
> *Subject:* Re: [Hps-analysis] HPS Analysis meeting March 5 @ 9am/noon 
> PST/EST
> After hearing Matt S. talk on Machine Learning, I realized there is a 
> big problem in MC production. To train Machine Learning, you need a 
> huge statistics of MC sample, especially if you want to train in each 
> mass bin. Furthermore, the MC sample should have beam-background 
> overlaid;  it should be tritrig-wab-beam not 
> tritrig-without-wab-beam.  A high statistics 1.05 GeV tritrig-wab-beam 
> sample with roughly equivalent to 2015 data statistics was generated 
> last year.  It took about 3 weeks to just generate wab-beam background 
> and another week to generate tritrig-wab-beam recon files.  It 
> required 50 TB to store wab-beam.SLIC files. Since there were no 50 TB 
> space,  earlier wab-beam files were deleted as the tritrig-wab-beam 
> recon files were completed.  Since 2016 run is higher energy and 4 
> times higher current, it will take more CPU time and need more disk 
> space.  If we clean-up disk space, MC production with data equivalent 
> statistics could be doable, but significantly higher statistics (10x 
> data) is difficult.
>
> Takashi
>
> -----Original Message-----
> From: Hps-analysis [mailto:hps-analysis-bounces at jlab.org] On Behalf Of 
> Graham, Mathew Thomas
> Sent: Tuesday, March 05, 2019 5:55 AM
> To: hps-analysis at jlab.org
> Subject: [Hps-analysis] HPS Analysis meeting March 5 @ 9am/noon PST/EST
>
>
> Hi All,
>
> Meeting today, here’re the details:
>
>
> https://gcc01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fconfluence.slac.stanford.edu%2Fdisplay%2Fhpsg%2FMarch%2B5%252C%2B2019%2BMeeting&data=02%7C01%7Chps-analysis%40jlab.org%7C9965abff0c774d82722d08d6a26d9857%7Cb4d7ee1f4fb34f0690372b5b522042ab%7C1%7C1%7C636874988922073875&sdata=uCAdtBi2diea9QkdeywW9vEJ4IOlYy52RiLmrQCG3Ng%3D&reserved=0 
> <https://gcc01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fconfluence.slac.stanford.edu%2Fdisplay%2Fhpsg%2FMarch%2B5%252C%2B2019%2BMeeting&data=02%7C01%7Cstepanya%40jlab.org%7Cafefec390bf34e20daa308d6a2723c5b%7Cb4d7ee1f4fb34f0690372b5b522042ab%7C1%7C0%7C636875008855054886&sdata=fuxTR33IBdHrVqP2WULhFQnRrpgYY83ZJ79AzgiVrgI%3D&reserved=0> 
> <https://gcc01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fconfluence.slac.stanford.edu%2Fdisplay%2Fhpsg%2FMarch%2B5%252C%2B2019%2BMeeting&data=02%7C01%7Chps-analysis%40jlab.org%7C9965abff0c774d82722d08d6a26d9857%7Cb4d7ee1f4fb34f0690372b5b522042ab%7C1%7C0%7C636874988922073875&sdata=oL2kjbmK3dU0J9TZ1%2BHfQdqrWWTIco6Ht0Xmuj46vVQ%3D&reserved=0 
> <https://gcc01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fconfluence.slac.stanford.edu%2Fdisplay%2Fhpsg%2FMarch%2B5%252C%2B2019%2BMeeting&data=02%7C01%7Cstepanya%40jlab.org%7Cafefec390bf34e20daa308d6a2723c5b%7Cb4d7ee1f4fb34f0690372b5b522042ab%7C1%7C0%7C636875008855064891&sdata=Inm4SMPTeqPmoUzu7uN3WFKgQax8Dyy9VSAoIPOENXs%3D&reserved=0>> 
>
>
> *       Bluejeans
>
>         *       Browser: 
> https://gcc01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fbluejeans.com%2F722990411&data=02%7C01%7Chps-analysis%40jlab.org%7C9965abff0c774d82722d08d6a26d9857%7Cb4d7ee1f4fb34f0690372b5b522042ab%7C1%7C1%7C636874988922073875&sdata=j9mdjCZCihzPoXSRO57NwTR8MeaZkIULTpu5MMHi1L0%3D&reserved=0 
> <https://gcc01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fbluejeans.com%2F722990411&data=02%7C01%7Cstepanya%40jlab.org%7Cafefec390bf34e20daa308d6a2723c5b%7Cb4d7ee1f4fb34f0690372b5b522042ab%7C1%7C0%7C636875008855074904&sdata=6soqXHDvea25oglXtaIgzsd90qZISKKQWeLC%2Fk0Vfpk%3D&reserved=0> 
> <https://gcc01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fbluejeans.com%2F722990411&data=02%7C01%7Chps-analysis%40jlab.org%7C9965abff0c774d82722d08d6a26d9857%7Cb4d7ee1f4fb34f0690372b5b522042ab%7C1%7C0%7C636874988922073875&sdata=3BkYOcFjYrCULT%2FTl%2BfbiNKCyt0b0YWeBvTz8gENsaM%3D&reserved=0 
> <https://gcc01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fbluejeans.com%2F722990411&data=02%7C01%7Cstepanya%40jlab.org%7Cafefec390bf34e20daa308d6a2723c5b%7Cb4d7ee1f4fb34f0690372b5b522042ab%7C1%7C0%7C636875008855074904&sdata=6soqXHDvea25oglXtaIgzsd90qZISKKQWeLC%2Fk0Vfpk%3D&reserved=0>> 
>
>         *       Phone:  8882402560  722990411  ##  *4
>         *       Details: 
> https://gcc01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fconfluence.slac.stanford.edu%2Fdisplay%2Fhpsg%2FDAWG%2BMeetings&data=02%7C01%7Chps-analysis%40jlab.org%7C9965abff0c774d82722d08d6a26d9857%7Cb4d7ee1f4fb34f0690372b5b522042ab%7C1%7C1%7C636874988922073875&sdata=HENB40iU30pa9D0zazC9qvcNL2pULJbsDnWinI2p650%3D&reserved=0 
> <https://gcc01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fconfluence.slac.stanford.edu%2Fdisplay%2Fhpsg%2FDAWG%2BMeetings&data=02%7C01%7Cstepanya%40jlab.org%7Cafefec390bf34e20daa308d6a2723c5b%7Cb4d7ee1f4fb34f0690372b5b522042ab%7C1%7C0%7C636875008855084908&sdata=HpjGDz8KR2%2Fe02oFJTku6f2m74FCxqt0M9X4OaaxHjw%3D&reserved=0> 
> <https://gcc01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fconfluence.slac.stanford.edu%2Fdisplay%2Fhpsg%2FDAWG%2BMeetings&data=02%7C01%7Chps-analysis%40jlab.org%7C9965abff0c774d82722d08d6a26d9857%7Cb4d7ee1f4fb34f0690372b5b522042ab%7C1%7C0%7C636874988922083880&sdata=GO6coc86kSSWiCv0rjKLqMqSuelqJH7i26j0rcuXjNs%3D&reserved=0 
> <https://gcc01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fconfluence.slac.stanford.edu%2Fdisplay%2Fhpsg%2FDAWG%2BMeetings&data=02%7C01%7Cstepanya%40jlab.org%7Cafefec390bf34e20daa308d6a2723c5b%7Cb4d7ee1f4fb34f0690372b5b522042ab%7C1%7C0%7C636875008855094917&sdata=sdKDroLarMYVsdOelaD99aQr4FB5y3skKlr49jwt4u8%3D&reserved=0>> 
>
>
> *       Meeting Rooms
>
>         *       JLAB: F228
>         *       SLAC: Ballam
>
>
> Agenda
>
>
> *       Machine Learning in Vertexing Analysis – MattS
> *       Relative SVT-ECal alignment in 2016 Data – Norman
> *       Bugfixes in beamspot-constrained vertexing – MattG
>
>
>
> _______________________________________________
> Hps-analysis mailing list
> Hps-analysis at jlab.org
> https://mailman.jlab.org/mailman/listinfo/hps-analysis
>
>
> _______________________________________________
> Hps-analysis mailing list
> Hps-analysis at jlab.org
> https://mailman.jlab.org/mailman/listinfo/hps-analysis

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.jlab.org/pipermail/hps-analysis/attachments/20190306/31590014/attachment.html>