Joosep Pata comments

Results 115 comments of


                                            Joosep Pata

Do we still need to support the muon reco duplicate mitigation in PhysicsTools/PatUtils?

Right, these can't be used because they are relying on old, removed code. Likely the best solution would be to remove these modules.

reco::Candidate pt = -nan in some rare cases

Even though #39120 will mitigate it for photons, I could imagine there's a high likelihood a NaN PFCandidate will mess up also other things. Therefore, @cms-sw/pf-l2 @laurenhay it might be...

instability in the runMETCorrectionsAndUncertainties

type jetmet

improved loss function

https://github.com/FlorentinCDX/Fast-Approx-SW/blob/master/sw_approx.py

Sliced n-dimensional Wasserstein was added here: https://github.com/jpata/particleflow/pull/123. So far, it did not drastically change the performance. Other options listed above are still on the table.

Add citation metadata wtih CITATION.cff

Perhaps related to this issue: is there a recommended list of proceedings/papers to cite for uproot (also for awkward)? Should it be somewhere in the README?

Provide an efficient inference implementation using sparsification/quantization

adding @raj2022

Provide an efficient inference implementation using sparsification/quantization

Also related: https://github.com/jpata/particleflow/issues/315

implement simple JEC in our validation code

Is L1 actually done on top of PUPPI? Maybe it's not needed. https://indico.cern.ch/event/783781/contributions/3386619/attachments/1832738/3001906/HEP2019_JECs_MD.pdf

improved loss function

basically, swd, genjet loss, MET loss have been implemented