The classification was made with the following final selection of variables:
j_Deep, "DeepCSV"
DR_lj, "DeltaR(l-jet)"
Deta_lj, "DeltaEta(l-jet)"
DpT_metj, "DeltapT(MET-jet)"
DPhi_lj, "DeltaPhi(l-jet)"
Deta_lmet, "Deta_lmet(l-MET)"
"DpT_lj", "DpT_lj(l-jet)"
"DR_metj", "DR_metj"
"DPhi_metj", "DPhi_metj"
With the following linear correlations:
-> There are no cuts applied (yet)
BDT result
Was used the default configuration for the trees:
"!H:!V:NTrees=850:MinNodeSize=2.5%:MaxDepth=3:BoostType=AdaBoost:AdaBoostBeta=0.5:UseBaggedBoost:BaggedSampleFraction=0.5:SeparationType=GiniIndex:nCuts=20"
DNN CPU result
The architecture used was:
RELU|128,RELU|128,RELU|128,SOFTSIGN
The classification was made with the following final selection of variables:
With the following linear correlations:
-> There are no cuts applied (yet)
BDT result
Was used the default configuration for the trees: "!H:!V:NTrees=850:MinNodeSize=2.5%:MaxDepth=3:BoostType=AdaBoost:AdaBoostBeta=0.5:UseBaggedBoost:BaggedSampleFraction=0.5:SeparationType=GiniIndex:nCuts=20"
DNN CPU result
The architecture used was: RELU|128,RELU|128,RELU|128,SOFTSIGN
With the following training strategy:
"TrainingStrategy=LearningRate=1e-2," "ConvergenceSteps=30,BatchSize=256,TestRepetitions=1," "WeightDecay=1e-4,Regularization=None," "DropConfig=0.0+0.5+0.5+0.5"
And the following DNN options: "!H:V:ErrorStrategy=CROSSENTROPY:VarTransform=N:" "WeightInitialization=XAVIER"
ROC curves: