[PWGJE] Add histograms for data-driven methods #8965
Conversation
hahassan7
requested review from
alibuild,
maoyx,
nzardosh,
fjonasALICE,
mfasDa and
mhemmer-cern
as code owners
|
Hi Hadi, why not use a table? |
|
Error while checking build/O2Physics/o2 for f9ced2e at 2024-12-12 17:08: Full log here. |
|
Hi Nima, tables might be extremely large. I want to minimize the output file as much as possible. |
|
Error while checking build/O2Physics/o2 for c570115 at 2024-12-12 17:21: Full log here. |
|
have you checked how large they would be? it seems you already have most of this information in trees anyways? |
|
The Trees are in the TreeCreator, there are no trees here. I also think THnSparse since it has binnings it will be much smaller than TTrees. |
|
ok fair enough |
|
At some point we should arrange a discussion on zoom about this HFtagging code and how to organise it. In general to me it seems the better way to organise it would be something like in HF where an ML selector task is ran which produces a tree with one score coloumn that can be joined to the jet table. Then this can be fed to the other tasks as a selection |
|
Yes, I totally agree with this. It will make tagging much easier with just taking the score from the extended jet table. I think we edit the one from Hanseo and add one score colunm to it and a function that evaluates the model and gives the score it gives. Though Hanseo's code needs so much cleaning. |
|
Ok that sounds great then, lets go in that direction :) Yeah I agree the code needs alot of cleaning. I think this would be a good thing to spend a week or so on since the code needs to be used in many workflows. I had discussed this at some point also with Hanseo that for Hard probes I approved alot of code changes that were urgent but with the understanding that they would be tidied up after so its a good moment perhaps to do this. |
|
I will try to do this maybe during the Christmas break, or a little bit after it. |
|
Error while checking build/O2Physics/o2 for c35118b at 2024-12-12 17:59: Full log here. |
vkucera
changed the title
|
@nzardosh @hahassan7 If it's fine with you, I'm converting this to a draft to avoid unnecessary recompilations since this is still in progress. |
|
Hi Hadi, I think we decided on changing the strategy right? to have the output such that the ML score would be joinable to the the jet table instead of how it is done here. |
Adding a THnSparse with jetpT score, jet mass, SV mass, and SVfE histograms wich will be used for data-driven approaches for efficiency and purity estimation.