CMU 4b Chats

• AE: low pt / Snakemake
• SM: getting ready to look at scalar samples
• CL: boosted

• AE: tightening the btagging (on all jets) doesnt help the lowpT … what if it is only on the pt3?
• LZ: Working on the boosted skim
• Francis: plans to run the SvB

• AE: some progress on low pt… still no change in btagging
• SM: Nothing yet
• JA: classifier studies

• AE: Low pT analysis: slow progress … 13 has not budged
• F : Getting started
• LZ: Slides for H
• SM: Starting

• AE: Low Pt Discussion
• LZ: Discussion of next steps

• AE: pushing on low pt
• SM: implented trigger wieghts
• CG: has CI going … will not make 24 friend trees
• LZ: needs clustering scripts from me

• AE: Low pt discussiosn/ HAs reana running on falcon
• LZ: Going on SV
• CL: Working on code clean up
• SM: Trigger debugging
• CGB; Wiill start with CIs

• AE: Looking more into low pt
• SM: Not much
• LZ: Will try to add the inference call to his processor
• (CGB: Should do Run4 CI with no FvT)

• JA: debugging FeynNet
• CL: will to look into span out clean up on coffea4bee code adding more benchmark plots MR to coffea about random number seeds
• AE: Memory is now monitored
• https://lss.fnal.gov/archive/test-tm/2000/fermilab-tm-2286-cd.pdf

• AE: SvB
• SM: debugging turn ons
• CL: updatong

• AE: away
• CL: MR almost ready… Will fix container issue
• Trigger eff: for 24

• AE: Following up on low pT
• CL: working on tests
• SM: Has skims
• skip rest b/c of Colloq

• Low pT retraining makes things worse !
• CL: pushed one set of checks working on another
• SM: getting skims

• AE: Getting there on low pT there is a bug in the low-pT FvT
• JA: MvD fits and issue-bot
• CL: working on tests
• SM: Working on lepton+ jets skims
• CG: working on 2024 skims
• LZ: understanding masses

• AE: low pt almost there
• JA: issue-bot / MvD fits
• SM: triggers
• CGB: skimming
• CL: ?
• LZ: boosted synthetic

AE: Low pt stuff / Falcon / Looking into Boosted settings JA: FSR Study / SvB classifier / CL: Working on tests for classier SM: No major updates / trigger selection CG: 24 TTToSemileptonic LZ::

• Monthly
• [ ] reAna
• [ ] Automatically make inputs / SvB
• [ ] FvT Friend trees ?

• JA: wrapping up mixing will move back to synthetic data
• AE: Will start doing fits
• SM: Progress on TRigger SFs
• CL: Working on CI
• CGB: Has skimms running will do one dataset
• ZH: Looking into cluster

• AE: JCM woes
• CL: updating the container and docs for the classifier
• SM: Skimming ttbar
• LZ: going to merge master and look into runing on PF cands

• Lei Zhao: Presentation of Boosted Synthetic
• SM: Went over her skimmer 1k lines long TTBar / Data skims : 4-jets / 2-jet + lepton
• AE: Working low pT anlaysis
• CL: Going to do some classifier docs

• AE: Starting to think about Low pt again

• AE: finding problems with anamolously large weights
• SM: Looking into the trigger turn ons.

AE: VBF puzzle. expecting 103 but see 75, think its because of change in systematic corelation AE to give approval talk on Friday SM: Run3 HH4b SvB looks very different from Run2 HH4b. Need to compare

AE: Close to having the VBF results SM: Has skims

• VBF samples done…
• Low pt the next step
• Sindhu to start with the trigger studies

• Need AUX material captions

• AE: redoing boosted skims / Low pt
• JA: Synthetic data
• CL: Not much … will send documentation
• SM: Working on closure will make an SvB cut

• SM: NEEDS TO MAKE SLIDES!!!!
• AE: understanding JCM code

• JA: working Cut flow
• AE: Starting low pt
• CL: Getting FvT and SvB running on new skims
• SM: working double peak

• AE Debugging skims
• Send thesis
• JA: Bootstrapping looks good / on to cutflow debugging
• SM: Still not getting it

• AE: low pt selection
• SM: memory errors
• CL: not much
• JA: new AN and synthetic data

• Virtual
• JA: working on the closure test with the Run2 synthetic dataset, making the Run 3 synthetic dataset, and answering the L2 comments.  - The Run2 synthetic closure looks pretty good now, Im still seeing some large FvT weights. Not clear where they are coming from. Chuyan, can you point to me where the event selection is done for the FvT training code ?  - The Run3 synthetic data looks good on our side, but Marina still sees big data vs syn-data differences. Im going to debug this with her.  - The L2 comments are mostly done. We still need the 2D plots (ΔR vs mass / and nViews vs Mass) – which I think AE is working on – and the Acc X Eff pltos which I am working on. Seems like the trigger weights are not being applied in the cut flow currently, 
• AE: Finally, fixed the binning on SvB., I now have to rerun everything (including the boosted combination), but the CERN REANA cluster has been very busy lately. I am figuring it out where to run the jobs and once they are done I need to update all the plots in the AN.
  • I need to do a new processor for the 2D plots, and the validation plots with the dileptonic ttbar selection. I am currently working on it.
  • On the CMU REANA side, I deployed REANA in our rogue02 and I have been testing it with the test workflows. Everything seems in place, but now I have to figure it out how to include CVFMS, users, accounts, storage area for processing jobs, etc. If I figure it out on time, I might try to run our analysis workflow there. Fingers crossed.
• CL: I am mainly improving the friend tree code to make it easier to use with new coffea and adding a dask-style configuration for the new processor and base classes.
• SM: reran the analysis with different jet pairings to see if it has an effect on the double peak. I'm rerunning that with just the hh region and its complement in the passDijetMass space. Also ran it with m4j cuts. Apart from that, I've been polishing up some of my slides and creating a writeup in prep for B2G.

• JA: working synthetic data and L2 plots
• AE: rebinning to get the signal flat / will look into redoing 2D plots
• CL: making plots from classifier friend trees

• AE to make the signal distribution flat, finishing L2 comments
• JA: needs to debug Synthetic data phase space
• CL: finishing background fits
• SM: putting slides together

• AE working comments
• CL: fitting new background method
• SM: not much

• CL some nice SvB Studies … Working on new background fit
• AE: Still dont understand klambda fits…<font color=red> NOW SOLVED!!! </font>
• JA: will make 15x synthetic data and try to run the closure
• Will get instructions on friend trees from AE and fitting FvT from CL

• SvB gains from adding other signal k_l points looks real but small.
• Will check stat only… if promising compute systematics
• In parallel CL, will refit FvT x 15 kfolds.

• AE: will make friend trees
• AE: updating the AN
• AE: ZZ and ZH fits
• AE: mixed data instead of data for v4.
• JA: Add CL stuff
• CL: More classifier FvT variations

• Debugging with CL
• FOUND THE pseudoJet Bug !! Very subtle.. see daily log above

• Lots of progress… no Sindhu [>>] Add ZZ/ZH studies to slides
• HH4b Run2 check consistency of years
• CL good progress on classifier studies: 20 min with data and ttbar
• AE has varible binning fits going

• Discussed answers to conveeners questions
• Talk in two weeks. [>>] John to prepare answers comments [>>] CL to send new SvB files .pkl to AE
• CL added singularity instructions to README

• JA
  • send the pickle file with ZZ in the SB
  • Setting up Falcon
• AE:
  • Debug limit change: trigger weights
  • Low pt: hours to make histograms double
• CL:
  • Finished classifier code
  • FvT with and without JCM
  • Send falcon instructions + data location
• SM:
  • Will start putting slides together

• CL: Classifier bug: cache sharing clashes.
• AE: close on low pt
• SM: debugging double peak

• AE: Will look into low pt: 3b data with JCM fit to signal in low Pt Marina has Run-3 ntuples to try Processor with simple run-3 selection
• CL: debugging the evaluation Plot tools and documentation Evaluation code Normalization issue
• JA: Look into boosted synthretic data Run-3 skims

• SM: Sculpting likely due to nJet and trigger. Will pursue new regions
• CH : Plotting tool / MC samples have the wrong normalization: AE to debug. Evaluation part of the classifier.
• AE: Finishing skims. Once finished will pursue low pt
• JA: Building to FvT fits

• CL: Most parts of the kappa scan plots work fine except for the stacked hists (I am using another plot library so it works in a little bit different way). I am trying to figure it out and I need to resolve some conflicts before merge. I noticed some hists e.g. nPVs, nPVGood, … are using Regular axis, maybe Integer axis works better which can be achieved by replacing 

  (101, -0.5, 100.5, (name, label)) with (0, 101, (name, label)).

  I also started testing the evaluation code, but it is not ready yet.

• AE: Has new skims, will make friend trees with trig weights and ttbar reco
• JA: Pushing on synthetic datasets. Need to think about residual mismodelling
• SM: Not much progress…

• AE: trigger troubles … JA look into converting c++ to python
• SM: Working on X→HH code going. Will look into new scheme for regions
• CL: Debugging K_l scan and writting out of the classifier weights
• JA: will look into converting c++ to python

• No sindhu
• JA: Long discussion of JetDeclustering
• AE: getting low pt going
• CL working on kl fits ?