Common interface discussion

[jpata]

Thanks Karim & Matthias for setting up the repo.

I think the format you propose is very good. For the b-tag likelihood ratio + MEM, the inputs you provide are sufficient. Here are some comments.

1. We should then also fix what are selectedJets, looseSelectedJets and selectedLeptons.
2. When we need to attach more information to a jet than just the 4-momentum and CSV, I agree we can make some very light classes for them.
3. We might want the GetOutput to return some class or struct which contains additional info, e.g. for the MEM you might like to save the integration accuracy and number of permutations for debugging purposes.
4. We might also like to think about how we include the boosted objects. In the MEM analysis we merge the resolved jets and subjets of boosted fatjets in case resolved jets do not satisfy the final state, however we have to think a bit about what's the best way to provide this procedure in a self-contained way.

[odysei]

Great to see this happening!

First of all, I strongly agree with Jooseph's 3rd point. However, I would take it one step further 3 structs: input, output, parameters.

In addition, I think we need a wrapper on top of this. I will push a suggestion in the next couple of days.

Lastly, let's not forget to start using a common coding style. For example, I made an early suggestion:
https://twiki.cern.ch/twiki/bin/viewauth/CMS/HiggsTopCodingGuide

[kelmorab]

Hi,

1. for the calculation of the BDT outputs the selected and loose objects are similar to those defined during our synchronization exercises.
2. I agree
3. Defining classes to hold the inputs, outputs and such is a good idea
4. Regarding the boosted objects i am not sure how we could handle those. For the first iteration of this framework i propose to neglect the boosted analysis. I will however point our boosted specialist to this discussion.
5. Regarding the wrapper i have no strong opinion about the architecture of this framework.
6. Thanks for pointing out this coding conventions, i will have a look at them.

[jpata]

Hi,

1. what are loose jets in the reference analysis?
2. OK, we will make a jet class when we will need more info on the jets
3. Yes, and lets try to keep the code(input, output structures) as minimal as possible (plain old data)!
4. OK to neglect boosted for now. We should make an effort to get both discriminators running and evaluating performances by next Wednesday, imho this is the top priority for this exercise.
5. meta-wrapper:
   Given that this repo will most likely contain relatively little actual code, but rather a light wrapper or interface to our separate projects, I propose to not focus too much right now on creating the appropriate structures, coding conventions etc before we have actually had a chance to iterate at least a few times to have some concrete comparison with bottlenecks on our hands. The aim is not to make a new, common analysis code, at least not at this stage. As long as the code here is minimal, the interference will be minimal and we can move fast.
6. conventions: as above, currently this is not the top priority before the checks are running.
7. One thing to discuss is the location of this package in the CMSSW source tree.
   I propose
   $CMSSW_BASE/src/TTH/CommonClassifier but you're welcome to change, I have no strong preference. This is needed to fix the include paths everywhere.
8. We also need to provide an install script for the actual MEM code, which lives in a separate git repo which we don't want to merge here in order to be able to benefit from updates.

Please see: #3 for 7-8.

I will provide a first functional wrapper from the ETH side by Monday evening at the latest.

[kelmorab]

Hi,

1. I am sorry, i thought it was stated on our synchronization twiki. The loose jets are the jets with a pT>20 and abs(eta)<2.4 in contrast to the other jets with pT>30 and abs(eta)<2.4.

[jpata]

As far as I know in the single-lepton channel we didn't actually define loose jets, in dilepton however we have

jet.pt() > 30 for all jets in single lepton, for leading 2 jets in dileptons
jet.pt() > 20 for all jets in dileptons, except 2 leading jets

So, does the BDT analysis also include jets of lower pt for the single-lepton leg?

[kelmorab]

If i am not mistaken in the current version of BDTs, in the ==5jet & >=4 b-tag category, the ttbar+H system is reconstructed by considering the highest pT<30GeV jet of this loose jet collection.

[jpata]

The btag-likelihood + MEM should now be testable by other groups.
I've also added a small test file root/events.json and a python script test/mem.py which calculate it.

Please let me know if you have troubles with setup.