This MR is to add the PFlow anti-kT 0.8 RC jets to TOPQ1. These jets are the base for the double-flavor tagger (DeXTer) being developed to be used in H->aa->bbXX searches. In the past, the jets themselves were not added to the derivation, but that created an inconsistency between training and application. Even though the inconsistency was eventually circumvented, it was really never solved. We would like to be able to add the proper jets to the derivation.
In order to ensure consistency, the code that produces the reclustered jets are now imported directly from FTAG derivation (as I think other large-R jet collections in TOPQ derivations do)
I am putting a WIP since it is important that the person training DeXTer check the code. @yuchou could you please check this MR? I put a small sample here http://rcoelhol.web.cern.ch/rcoelhol/TOPQ1_RCJet08/ . It contains the Reco_tf.py logs, and the content of the DAOD. I have also ran FTAG1 over the same AOD for direct 1-1 comparison. In particular, I am only saving the basic variables, but you should let me know if you need any extra variable (Mazin used a lot sd0 of the jets as an observable in the analysis, so this would be a natural candidate)
@spalazzo @borisov : Could you please let me know what you think about this proposal? I kept the variables we used in the 36 ifb analysis https://gitlab.cern.ch/atlas/athena/-/blob/21.2/PhysicsAnalysis/DerivationFramework/DerivationFrameworkTop/python/TOPQCommonExtraContent.py#L171-224 but I honestly believe no one else is using them. We won't need them anymore. We would greatly appreciate if we could include the jets that we use for flavor tagging. Please, let me know what you think and if there is some information you would need. The plan is to also include the flavor tagging discriminants and jet mass corrections. @yuchou , in time, will add to this MR with the appropriate modifications.
For reference, this is the JIRA of the previous strategy https://its.cern.ch/jira/browse/TOPQDERIV-62 that lead to the issues mentioned above.