Index: trunk/Sample_signal.root
===================================================================
Cannot display: file marked as a binary type.
svn:mime-type = application/octet-stream
Index: trunk/Sample_signal.root
===================================================================
--- trunk/Sample_signal.root (revision 1)
+++ trunk/Sample_signal.root (revision 2)
Property changes on: trunk/Sample_signal.root
___________________________________________________________________
Added: svn:mime-type
## -0,0 +1 ##
+application/octet-stream
\ No newline at end of property
Index: trunk/Sample_background.root
===================================================================
Cannot display: file marked as a binary type.
svn:mime-type = application/octet-stream
Index: trunk/Sample_background.root
===================================================================
--- trunk/Sample_background.root (revision 1)
+++ trunk/Sample_background.root (revision 2)
Property changes on: trunk/Sample_background.root
___________________________________________________________________
Added: svn:mime-type
## -0,0 +1 ##
+application/octet-stream
\ No newline at end of property
Index: trunk/TreeCreator.py
===================================================================
--- trunk/TreeCreator.py (revision 0)
+++ trunk/TreeCreator.py (revision 2)
@@ -0,0 +1,52 @@
+import sys
+import ROOT
+import ObsDict
+from array import array
+
+def processTrees( names = ["vPx", "vPy", "vPz", "vE"] ):
+ print "=======> Initializing TTree reprocessor..."
+ print "=======> Reading branches " + str(names) + "..."
+
+ #Opening the files that will be reprocessed
+ #These files contain only four branches, indicated in "names" list
+ Signal = ROOT.TFile(sys.argv[1], "read")
+ Background = ROOT.TFile(sys.argv[2], "read")
+ Data = ROOT.TFile(sys.argv[3], "read")
+ fileList = {Signal:"Signal", Background:"Background", Data:"Data"}
+
+ print "=======> Creating input TTrees..."
+ #Creating new files, that will be used for classification
+ #These files will contain new branches, which correspond to the observables indicated in ObsDict.py
+ fileN = 1
+ for file in fileList: #Looping over the files (signal, background and data)
+ print "=======> File " + sys.argv[fileN] + " is being reproduced..."
+ outputFile = ROOT.TFile("Reproc_" + fileList[file] + ".root", "recreate")
+ inputJetsTree = file.Get("jetsTree")
+ inputBosonTree = file.Get("bosonTree")
+ tree = ROOT.TTree("observablesTree", "observablesTree")
+ varN = 0
+ for var in ObsDict.obsList: #Looping over the observables we want to construct (see ObsDict.py)
+ print " Variable " + str(ObsDict.obsList[varN]) + " is calculating..."
+ value = array('f', [0.]) #Creating a branch variable
+ tree.Branch(var, value, var + "/F") #Creating the branch itself
+ if inputJetsTree.GetEntries() != inputBosonTree.GetEntries():
+ raise ValueError("The number of events is different in jets and boson TTrees! Check consistency of the samples!")
+ for i in range(inputJetsTree.GetEntries()):
+ #for events in inputJetsTree: #Looping over the events in input trees
+ inputJetsTree.GetEntry(i)
+ jet1 = ROOT.TLorentzVector() #Defining three TLorentzVector jets and
+ jet2 = ROOT.TLorentzVector() #filling them from the input trees
+ jet1.SetPxPyPzE(inputJetsTree.vPx[0], inputJetsTree.vPy[0], inputJetsTree.vPz[0], inputJetsTree.vE[0])
+ jet2.SetPxPyPzE(inputJetsTree.vPx[1], inputJetsTree.vPy[1], inputJetsTree.vPz[1], inputJetsTree.vE[1])
+ if ( jet1.E() == - jet2.E() ) or ( jet1.Px() == jet2.Px() and jet1.Py() == jet2.Py() ) or ( jet1.Eta() == jet2.Eta() ): continue
+ inputBosonTree.GetEntry(i)
+ boson = ROOT.TLorentzVector()
+ boson.SetPxPyPzE(inputBosonTree.vPx[0], inputBosonTree.vPy[0], inputBosonTree.vPz[0], inputBosonTree.vE[0])
+ value[0] = ObsDict.observables(jet1, jet2, boson, var) #Taking the jets' four-vectors and calculating the observable
+ tree.Fill() #Filling the event in new tree
+ varN += 1
+ outputFile.Write() #Writing the branch in tree
+ outputFile.Close() #Close the output file
+ print "=======> Done"
+ fileN += 1
+ print "=======> Input TTees have been created!"
\ No newline at end of file
Index: trunk/Classifier.py
===================================================================
--- trunk/Classifier.py (revision 0)
+++ trunk/Classifier.py (revision 2)
@@ -0,0 +1,93 @@
+import sys
+import ROOT
+import ObsDict
+import TreeCreator
+from array import array
+from ROOT import gROOT, TMVA, TMath, TFile, TLorentzVector
+
+#Main function of the classifier.
+def main():
+ #Check that user has indicated filenames
+ if sys.argv[3] is "":
+ raise ValueError("You must specify signal, background and data filenames!")
+
+ #If user provided False as a fourth argument, TTree reprocessing will not be performed
+ if len(sys.argv) == 5:
+ if sys.argv[4] == "False":
+ pass
+ elif sys.argv[4] == "True":
+ TreeCreator.processTrees() #This function is launching TTree reprocessing, i.e. Px, Py, Pz, E -> mjj, pTjj, ...
+ else:
+ raise ValueError("You must specify the correct reprocessing flag (True\False)")
+ else:
+ raise ValueError("You must specify the reprocessing flag at the end of the command line (True/False)")
+
+ print "=======> Launching BDT classifier..."
+
+ #The definition of reprocessed input TTrees
+ trainSignal = ROOT.TFile("Reproc_Signal.root", "read") #VBF
+ trainBackgr = ROOT.TFile("Reproc_Background.root", "read") #ggF
+ Data = ROOT.TFile("Reproc_Data.root", "read")
+
+ #Define the output file
+ outputFile = ROOT.TFile("Output_BDT.root", "recreate")
+
+ #Taking the TTrees from the input files
+ treeSignal = trainSignal.Get("observablesTree")
+ treeBackgr = trainBackgr.Get("observablesTree")
+ treeData = Data.Get("observablesTree")
+
+ #Define the TMVA factory object
+ factory = TMVA.Factory("TMVAJob", outputFile)
+
+ #Signal and background weights that will be used for training
+ signalWeight = 1.0
+ backgrWeight = 1.0
+
+ #Adding signal, background and regression trees in the factory
+ factory.AddSignalTree(treeSignal, signalWeight)
+ factory.AddBackgroundTree(treeBackgr, backgrWeight)
+ factory.AddRegressionTree(treeData, signalWeight)
+
+ #Tell the factory how to use the training and testing events
+ #factory.PrepareTrainingAndTestTree(mycuts, mycutb, "nTrain_Signal=0:nTrain_Background=0:SplitMode=Random:NormMode=NumEvents:!V")
+
+ #Including all the variables that we indicated in ObsDict.py
+ for var in ObsDict.obsList:
+ factory.AddVariable(var, 'F')
+ factory.AddTarget(var, 'F')
+
+ #Apply additional cuts on the signal and background samples (can be different)
+ mycuts = ""
+ mycutb = ""
+
+ #Booking the classifier method, it is possible to book several of them at once
+ factory.BookMethod( TMVA.Types.kBDT, "BDTG",
+ "!H:!V:NTrees=1000:BoostType=Grad:Shrinkage=0.30:UseBaggedBoost:GradBaggingFraction=0.6:SeparationType=GiniIndex:nCuts=20:PruneMethod=CostComplexity:PruneStrength=50:MaxDepth=2" )
+
+ #Performing a training, testing and evaluation of all methods
+ factory.TrainAllMethods()
+ factory.TestAllMethods()
+ factory.EvaluateAllMethods();
+
+ outputFile.Close();
+
+ print "=======> Wrote root file: " + outputFile.GetName()
+ print "=======> TMVAClassificationCategory is done!"
+
+ #Launching multi-class an regression GUI interfaces to check the plots
+ if not gROOT.IsBatch():
+ TMVA.TMVAMultiClassGui(outputFile.GetName())
+ TMVA.TMVARegGui(outputFile.GetName())
+
+ if not gROOT.IsBatch():
+ TMVA.TMVARegGui(outputFile.GetName())
+ TMVA.TMVAMultiClassGui(outputFile.GetName())
+
+ input("=======> Opening GUI interface. Press any key to continue.")
+ print "=======> Done!"
+
+ return 0
+
+if __name__ == "__main__":
+ main()
Index: trunk/Sample_data.root
===================================================================
Cannot display: file marked as a binary type.
svn:mime-type = application/octet-stream
Index: trunk/Sample_data.root
===================================================================
--- trunk/Sample_data.root (revision 1)
+++ trunk/Sample_data.root (revision 2)
Property changes on: trunk/Sample_data.root
___________________________________________________________________
Added: svn:mime-type
## -0,0 +1 ##
+application/octet-stream
\ No newline at end of property
Index: trunk/ObsDict.py
===================================================================
--- trunk/ObsDict.py (revision 0)
+++ trunk/ObsDict.py (revision 2)
@@ -0,0 +1,25 @@
+import ROOT
+from ROOT import TMath
+
+#Names of the observables we want to have in the reprocessed files
+obsList = ("pTBalance", "Mjj", "dEtajj", "dRjj", "Sin0p5dFi", "EtaCentrality", "pTCentrality", "Ejets")
+
+#A function, that take two jets' four-vectors as an input and
+#calculates the observable based on the provided key (obsName)
+def observables( jet1, jet2, boson, obsName ):
+ funcObsList = {
+ obsList[0]:TMath.Sqrt( ( jet1.Px() + jet2.Px() + boson.Px() )*( jet1.Px() + jet2.Px() + boson.Px() ) +
+ ( jet1.Py() + jet2.Py() + boson.Py() )*( jet1.Py() + jet2.Py() + boson.Py() ) )/( jet1.Pt() + jet2.Pt() + boson.Pt() ),
+ obsList[1]:(jet1 + jet2).M(),
+ obsList[2]:TMath.Abs( jet1.Eta() - jet2.Eta() ),
+ obsList[3]:TMath.Sqrt( ( jet1.Eta() - jet2.Eta() )*( jet1.Eta() - jet2.Eta() ) +
+ ( jet1.Phi() - jet2.Phi() )*( jet1.Phi() - jet2.Phi() ) ),
+ obsList[4]:TMath.Sin( TMath.Abs( ( jet1.Phi() - jet2.Phi() )/2. ) ),
+ obsList[5]:TMath.Abs( ( boson.Eta() - ( jet1.Eta() - jet2.Eta() )/2. )/( jet1.Eta() - jet2.Eta() ) ),
+ obsList[6]:TMath.Sqrt( ( boson.Px() - jet1.Px()/2. - jet1.Px()/2. )*( boson.Px() - jet1.Px()/2. - jet1.Px()/2. ) +
+ ( boson.Py() - jet1.Py()/2. - jet1.Py()/2. )*( boson.Py() - jet1.Py()/2. - jet1.Py()/2. ) )/TMath.Sqrt( ( jet1.Px() - jet2.Px() )*( jet1.Px() - jet2.Px() ) +
+ ( jet1.Py() - jet2.Py() )*( jet1.Py() - jet2.Py() ) ),
+ obsList[7]:( jet1.Pt() + jet2.Pt() )/( jet1.E() + jet2.E() )
+ }
+ if obsName in obsList: return funcObsList[obsList[obsList.index(obsName)]]
+ else: print "Variable " + ''.join(obsName) + " is not defined!"
\ No newline at end of file