[Hadstruct] [EXTERNAL] Further work with NNPDF

[Joseph Karpie]

The distillation results were my previous example, because those look so
much better than anything else we've had. That makes me want to study
them. There are a number of other things that come to mind which we can
begin studying which do not require new unpublished data. Examples are a
"global" combined analysis of all our pion results which go straight from
matrix elements to PDFs or the effects of cuts on the data ranges on final
PDFs as Tommaso suggests.

On our call, I think we should discuss what, if any, of these things that
we would like to do with them and what we would like to do ourselves.

Best,
Joe

On Tue, Sep 22, 2020 at 1:26 PM Savvas Zafeiropoulos <savvaslz at gmail.com>
wrote:

> Yeah absolutely! Nobody suggested sharing our pristine distillation
> results with non-hadstruc people before publication!
> Of course it would be better to discuss this live in our zoom call...
>
> Regards,
> Savvas
>
> On Tue, Sep 22, 2020 at 8:17 PM Joseph Karpie <jmkarpie at email.wm.edu>
> wrote:
>
>> Savvas is right that when I said "when they are finalized" I was thinking
>> after publication.
>>
>> On Tue, Sep 22, 2020 at 1:06 PM Savvas Zafeiropoulos <savvaslz at gmail.com>
>> wrote:
>>
>>> I think Joe's suggestion is to do something with NNPDF after we publish
>>> our distillation data alone as hadstruc...
>>> From both the heavy and lighter pion masses...
>>>
>>> On Tue, Sep 22, 2020 at 8:03 PM Konstantinos Orginos <kostas at jlab.org>
>>> wrote:
>>>
>>>> In my view we should first let Colin write the paper publish it and
>>>> then discuss further collaboration with NNPDF.
>>>> By  adding new authors to a paper that is nearly complete is not in the
>>>> best interests of  our own graduates students who  worked hard on this
>>>> project
>>>> and deserve a first take in analysing and publishing the data without
>>>> external interference.
>>>>
>>>> Cheers,
>>>>
>>>> Kostas.
>>>>
>>>>
>>>> > On Sep 22, 2020, at 1:56 PM, Joseph Karpie <jkarpie at gmail.com> wrote:
>>>> >
>>>> > Hey everyone,
>>>> >
>>>> > As some of you may know, a few of us are wrapping up a paper with
>>>> Luigi Del Debbio and Tommaso Giana (NNPDF people) where they reanalyse our
>>>> Gaussian smeared results using their tools. Tommaso sent around a number of
>>>> suggestions for things to begin studying next, such as studying fit forms
>>>> and using only high or low Ioffe time data.
>>>> >
>>>> > They have experience using sophisticated methods to perform the
>>>> inverse and ways of accounting for systematic errors which can be very
>>>> useful to us. I think we should discuss Friday what future work we would
>>>> like to be doing with them.
>>>> >
>>>> > For example, I'd like to do a comparison of our Gaussian smeared
>>>> results with the distillation results that have been made on the same
>>>> ensemble when they are finalized. My thought is that the data are more or
>>>> less consistent, but Colin's fits appear to have significantly better
>>>> agreement with pheno results in the x > 0.5 region than my own fits or
>>>> Tommaso's NNPDF fits to the Gaussian smeared data. I want to understand
>>>> what difference between the results created this dramatic improvement. I
>>>> believe working with NNPDF would make this task significantly easier due to
>>>> their experience and tools.
>>>> >
>>>> > Best,
>>>> > Joe
>>>> > _______________________________________________
>>>> > Hadstruct mailing list
>>>> > Hadstruct at jlab.org
>>>> > https://mailman.jlab.org/mailman/listinfo/hadstruct
>>>>
>>>>
>>>> _______________________________________________
>>>> Hadstruct mailing list
>>>> Hadstruct at jlab.org
>>>> https://mailman.jlab.org/mailman/listinfo/hadstruct
>>>>
>>> _______________________________________________
>>> Hadstruct mailing list
>>> Hadstruct at jlab.org
>>> https://mailman.jlab.org/mailman/listinfo/hadstruct
>>>
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.jlab.org/pipermail/hadstruct/attachments/20200922/b660d9df/attachment-0001.html>