Repository navigation
Bdt asimov opt - #18
elisejsledge wants to merge 20 commits into
Conversation
MatBarria
left a comment
There was a problem hiding this comment.
I think it looks great, just one general comment. Are we using the best version of the BDT by default in all the codes?. I think if I just run everything that version should be use.
There was a problem hiding this comment.
uncomment list_datasets so it runs over all the datasets
There was a problem hiding this comment.
uncomment list_datasets so it runs over all the datasets
| if ((n_bjet == 0) && (n_bjet_Loose < 2) && | ||
| (elec_pt->size() == 0) && (n_jet >= 2) && leading_jet_pt > 35 && | ||
| (diJet_mass >= 400) && (delta_eta_diJet > 2.5)) { | ||
| (diJet_mass >= 400) && (delta_eta_diJet > 2.5) && (ifSync && subleading_jet_pt > 30)) { |
There was a problem hiding this comment.
is this working without the ifSync?
| n_valid_muons++; | ||
| } | ||
| } | ||
| else if(ifSync && pt_raw > 20 && Muon_isGlobal[i] && |
There was a problem hiding this comment.
Maybe remove ifSync here, I think is little confusing because it doesn't change the logic
| EOS_DIR="/store/group/lpchmumu/${USER_NAME}/analyzer_HiggsMuMu_${ANALYZER_VERSION}/tuples_${TUPLES_VERSION}/systematics" | ||
| # ---- Hardcoded locations: edit these ---- | ||
| # Systematic tuples to score (/store/group/lpchmumu is the same area as /store/user/lpchmumu) | ||
| SYST_TUPLE_DIR="/store/group/lpchmumu/mbarrial/analyzer_HiggsMuMu_${ANALYZER_VERSION}/tuples_${TUPLES_VERSION}/systematics" |
There was a problem hiding this comment.
we have to unhardcode my path here I think, at least for the git version
There was a problem hiding this comment.
can we make the suffix optional here? we will just use one of them, right? I think this by the default should use the best version (asimov fit flat I think) and save it as allData_combined.root without suffix. That way we don't need to add the suffix in the other code too.
There was a problem hiding this comment.
can we make the suffix optional here? we will just use one of them, right? I think this by the default should use the best version (asimov fit flat I think) and save it as allData_combined.root without suffix. That way we don't need to add the suffix in the other code too.
| # train.py variant label of the BDT (e.g. "base", "asimovFit_flatMassRes"). It selects the | ||
| # <channel><suffix> entry of root_io/bdt_categories.json and the nominal BDT_score/<channel><suffix>/ | ||
| # folder, so it must match the model the systematic tuples were scored with. | ||
| BDT_VARIANT = "base" |
There was a problem hiding this comment.
is this the best version? can we use the best version by default?
Adding asimov optimization and k-fold cross-validation to the optuna hyperparameter optimization. Also, updating systematic scripts to account for new BDT_score variants and the change in xgboost version for the base cmssw python install