Hi all.
OK, so let’s aim to meet sometime within the next 2 weeks - as soon as tasks are completed.
Tasks:
1) Complete all MC production and data/MC comparisons for given configurations.
2) Complete table with qualitative characterisation of configurations, for all data/MC comparisons
we can run, separating out different kinematical regions etc in each comparison.
3) Add table to provide quantitative level of agreement, for all comparisons that provide this functionality.
4) Add all new plots.
We should allow 1-2 days to look at the plots and digest the information.
Then let’s meet to discuss:
1) Buyer’s remorse! Even without looking at own comparisons, do we have second thoughts about the tested configurations?
Are there alternative candidate configurations we want to test? What is the motivation?
2) Based on data/MC comparisons: Are current configurations performing as expected?
If we suspect a bug, outline plan for further investigation.
3) Based on the pattern of data/MC agreement, do we see opportunities to improve the configuration
(e.g by minimal tuning of an existing parameter, or by a minimal tuning of a new parameter that gives more freedom to
simultaneously “fit” multiple datasets, or by defining an alternative configuration by swapping some component models in/out).
I am hoping that the revised document / evidence provided will be improved substantially, but what is already available can
already help some thinking on items 1-3 above.
Also, hoping that stage 1 in Professor/GENIE tuning exercise would be completed so we can find out how it went.
cheers
Costas
> On 22 Nov 2016, at 12:08, Costas Andreopoulos <[log in to unmask]> wrote:
>
> Some material for the discussion today
>
> https://dl.dropboxusercontent.com/u/45977020/ProfGENIE_20161122.pdf
>
> https://dl.dropboxusercontent.com/u/45977020/GENIEv3_characterization_v1.0.pdf
>
> cheers
> Costas
>
> On 21 Nov 2016, at 23:35, Costas Andreopoulos <[log in to unmask]<mailto:[log in to unmask]>> wrote:
>
> Yes, of course we could do that, if someone is willing to invest the effort.
>
> Our core mission is to deliver improved versions of the generator and generator tunes.
> In comparison, supporting generator-related analysis tasks in experiments is secondary.
>
> The latter should not interfere with the former. So I would like to avoid resistance in including in our tunes parameters
> which are non-reweightable in principle (or parameters for which reweighing code has not been written yet) on the basis
> that uncertainties can not be propagated using event reweighing.
>
> cheers
> C
>
>
> On 21 Nov 2016, at 23:02, Gabriel Nathan Perdue <[log in to unmask]<mailto:[log in to unmask]><mailto:[log in to unmask]>> wrote:
>
>
> On Nov 21, 2016, at 2:47 PM, Costas Andreopoulos <[log in to unmask]<mailto:[log in to unmask]><mailto:[log in to unmask]>> wrote:
>
> The GENIE reweighting needs a major upgrade to properly initialise itself with configuration data stored in the event file it reweighs. Otherwise, it is all too easy for differences to creep in. I always wanted to do this upgrade, which is very substantial, but i do not see the motivation for this any more. Professor does not need reweighting, so our tunes won't depend critically upon it. So it is certain that there will be a increasingly widening gap between the amount of tuning and error estimation done in GENIE, and the amount of it that can be supported via reweighting. Users need to develop own solutions for their analysis needs.
>
> This is all true, but the re-weighting facilities are a premier feature in GENIE and we should strongly consider continuing to modernize and support them at a high level.
>
>
> Gabriel Perdue
> Associate Scientist
>
> Scientific Computing Division
> Fermi National Accelerator Laboratory
> PO Box 500, MS 234, Batavia, IL 60510, USA
> [log in to unmask]<mailto:[log in to unmask]><mailto:[log in to unmask]>
> Office: 630-840-6499
> Cell: 630-605-8062
>
> Connect with us!
> Newsletter<http://news.fnal.gov/fermilab-at-work/> | Facebook<http://www.facebook.com/Fermilab> | Twitter<https://www.twitter.com/fermilab>
|