#!/usr/bin/env python # @(#)root/tmva $Id$ # ------------------------------------------------------------------------------ # # Project : TMVA - a Root-integrated toolkit for multivariate data analysis # # Package : TMVA # # Python script: TMVAClassification.py # # # # This python script provides examples for the training and testing of all the # # TMVA classifiers through PyROOT. # # # # The Application works similarly, please see: # # TMVA/macros/TMVAClassificationApplication.C # # For regression, see: # # TMVA/macros/TMVARegression.C # # TMVA/macros/TMVARegressionpplication.C # # and translate to python as done here. # # # # As input data is used a toy-MC sample consisting of four Gaussian-distributed # # and linearly correlated input variables. # # # # The methods to be used can be switched on and off via the prompt command, for # # example: # # # # python TMVAClassification.py --methods Fisher,Likelihood # # # # The output file "TMVA.root" can be analysed with the use of dedicated # # macros (simply say: root -l <../macros/macro.C>), which can be conveniently # # invoked through a GUI that will appear at the end of the run of this macro. # # # # for help type "python TMVAClassification.py --help" # # ------------------------------------------------------------------------------ # # -------------------------------------------- # Standard python import import sys # exit import time # time accounting import getopt # command line parser # -------------------------------------------- # Default settings for command line arguments DEFAULT_OUTFNAME = "TMVA.root" DEFAULT_INFNAME = "tmva_class_example.root" DEFAULT_TREESIG = "TreeS" DEFAULT_TREEBKG = "TreeB" DEFAULT_METHODS = "Cuts,CutsD,CutsPCA,CutsGA,CutsSA,Likelihood,LikelihoodD,LikelihoodPCA,LikelihoodKDE,LikelihoodMIX,PDERS,PDERSD,PDERSPCA,PDEFoam,PDEFoamBoost,KNN,LD,Fisher,FisherG,BoostedFisher,HMatrix,FDA_GA,FDA_SA,FDA_MC,FDA_MT,FDA_GAMT,FDA_MCMT,MLP,MLPBFGS,MLPBNN,CFMlpANN,TMlpANN,SVM,BDT,BDTD,BDTG,BDTB,RuleFit" # Print usage help def usage(): print " " print "Usage: python %s [options]" % sys.argv[0] print " -m | --methods : gives methods to be run (default: all methods)" print " -i | --inputfile : name of input ROOT file (default: '%s')" % DEFAULT_INFNAME print " -o | --outputfile : name of output ROOT file containing results (default: '%s')" % DEFAULT_OUTFNAME print " -t | --inputtrees : input ROOT Trees for signal and background (default: '%s %s')" \ % (DEFAULT_TREESIG, DEFAULT_TREEBKG) print " -v | --verbose" print " -? | --usage : print this help message" print " -h | --help : print this help message" print " " # Main routine def main(): try: # retrive command line options shortopts = "m:i:t:o:vh?" longopts = ["methods=", "inputfile=", "inputtrees=", "outputfile=", "verbose", "help", "usage"] opts, args = getopt.getopt( sys.argv[1:], shortopts, longopts ) except getopt.GetoptError: # print help information and exit: print "ERROR: unknown options in argument %s" % sys.argv[1:] usage() sys.exit(1) infname = DEFAULT_INFNAME treeNameSig = DEFAULT_TREESIG treeNameBkg = DEFAULT_TREEBKG outfname = DEFAULT_OUTFNAME methods = DEFAULT_METHODS verbose = False for o, a in opts: if o in ("-?", "-h", "--help", "--usage"): usage() sys.exit(0) elif o in ("-m", "--methods"): methods = a elif o in ("-i", "--inputfile"): infname = a elif o in ("-o", "--outputfile"): outfname = a elif o in ("-t", "--inputtrees"): a.strip() trees = a.rsplit( ' ' ) trees.sort() trees.reverse() if len(trees)-trees.count('') != 2: print "ERROR: need to give two trees (each one for signal and background)" print trees sys.exit(1) treeNameSig = trees[0] treeNameBkg = trees[1] elif o in ("-v", "--verbose"): verbose = True # Print methods mlist = methods.replace(' ',',').split(',') print "=== TMVAClassification: use method(s)..." for m in mlist: if m.strip() != '': print "=== - <%s>" % m.strip() # Import ROOT classes from ROOT import gSystem, gROOT, gApplication, TFile, TTree, TCut # check ROOT version, give alarm if 5.18 if gROOT.GetVersionCode() >= 332288 and gROOT.GetVersionCode() < 332544: print "*** You are running ROOT version 5.18, which has problems in PyROOT such that TMVA" print "*** does not run properly (function calls with enums in the argument are ignored)." print "*** Solution: either use CINT or a C++ compiled version (see TMVA/macros or TMVA/examples)," print "*** or use another ROOT version (e.g., ROOT 5.19)." sys.exit(1) # Logon not automatically loaded through PyROOT (logon loads TMVA library) load also GUI gROOT.SetMacroPath( "./" ) gROOT.Macro ( "./TMVAlogon.C" ) gROOT.LoadMacro ( "./TMVAGui.C" ) # Import TMVA classes from ROOT from ROOT import TMVA # Output file outputFile = TFile( outfname, 'RECREATE' ) # Create instance of TMVA factory (see TMVA/macros/TMVAClassification.C for more factory options) # All TMVA output can be suppressed by removing the "!" (not) in # front of the "Silent" argument in the option string factory = TMVA.Factory( "TMVAClassification", outputFile, "!V:!Silent:Color:DrawProgressBar:Transformations=I;D;P;G,D:AnalysisType=Classification" ) # Set verbosity factory.SetVerbose( verbose ) # If you wish to modify default settings # (please check "src/Config.h" to see all available global options) # gConfig().GetVariablePlotting()).fTimesRMS = 8.0 # gConfig().GetIONames()).fWeightFileDir = "myWeightDirectory" # Define the input variables that shall be used for the classifier training # note that you may also use variable expressions, such as: "3*var1/var2*abs(var3)" # [all types of expressions that can also be parsed by TTree::Draw( "expression" )] factory.AddVariable( "myvar1 := var1+var2", 'F' ) factory.AddVariable( "myvar2 := var1-var2", "Expression 2", "", 'F' ) factory.AddVariable( "var3", "Variable 3", "units", 'F' ) factory.AddVariable( "var4", "Variable 4", "units", 'F' ) # You can add so-called "Spectator variables", which are not used in the MVA training, # but will appear in the final "TestTree" produced by TMVA. This TestTree will contain the # input variables, the response values of all trained MVAs, and the spectator variables factory.AddSpectator( "spec1:=var1*2", "Spectator 1", "units", 'F' ) factory.AddSpectator( "spec2:=var1*3", "Spectator 2", "units", 'F' ) # Read input data if gSystem.AccessPathName( infname ) != 0: gSystem.Exec( "wget http://root.cern.ch/files/" + infname ) input = TFile.Open( infname ) # Get the signal and background trees for training signal = input.Get( treeNameSig ) background = input.Get( treeNameBkg ) # Global event weights (see below for setting event-wise weights) signalWeight = 1.0 backgroundWeight = 1.0 # ====== register trees ==================================================== # # the following method is the prefered one: # you can add an arbitrary number of signal or background trees factory.AddSignalTree ( signal, signalWeight ) factory.AddBackgroundTree( background, backgroundWeight ) # To give different trees for training and testing, do as follows: # factory.AddSignalTree( signalTrainingTree, signalTrainWeight, "Training" ) # factory.AddSignalTree( signalTestTree, signalTestWeight, "Test" ) # Use the following code instead of the above two or four lines to add signal and background # training and test events "by hand" # NOTE that in this case one should not give expressions (such as "var1+var2") in the input # variable definition, but simply compute the expression before adding the event # # # --- begin ---------------------------------------------------------- # # ... *** please lookup code in TMVA/macros/TMVAClassification.C *** # # # --- end ------------------------------------------------------------ # # ====== end of register trees ============================================== # Set individual event weights (the variables must exist in the original TTree) # for signal : factory.SetSignalWeightExpression ("weight1*weight2"); # for background: factory.SetBackgroundWeightExpression("weight1*weight2"); factory.SetBackgroundWeightExpression( "weight" ) # Apply additional cuts on the signal and background sample. # example for cut: mycut = TCut( "abs(var1)<0.5 && abs(var2-0.5)<1" ) mycutSig = TCut( "" ) mycutBkg = TCut( "" ) # Here, the relevant variables are copied over in new, slim trees that are # used for TMVA training and testing # "SplitMode=Random" means that the input events are randomly shuffled before # splitting them into training and test samples factory.PrepareTrainingAndTestTree( mycutSig, mycutBkg, "nTrain_Signal=0:nTrain_Background=0:SplitMode=Random:NormMode=NumEvents:!V" ) # -------------------------------------------------------------------------------------------------- # ---- Book MVA methods # # please lookup the various method configuration options in the corresponding cxx files, eg: # src/MethoCuts.cxx, etc, or here: http://tmva.sourceforge.net/optionRef.html # it is possible to preset ranges in the option string in which the cut optimisation should be done: # "...:CutRangeMin[2]=-1:CutRangeMax[2]=1"...", where [2] is the third input variable # Cut optimisation if "Cuts" in mlist: factory.BookMethod( TMVA.Types.kCuts, "Cuts", "!H:!V:FitMethod=MC:EffSel:SampleSize=200000:VarProp=FSmart" ) if "CutsD" in mlist: factory.BookMethod( TMVA.Types.kCuts, "CutsD", "!H:!V:FitMethod=MC:EffSel:SampleSize=200000:VarProp=FSmart:VarTransform=Decorrelate" ) if "CutsPCA" in mlist: factory.BookMethod( TMVA.Types.kCuts, "CutsPCA", "!H:!V:FitMethod=MC:EffSel:SampleSize=200000:VarProp=FSmart:VarTransform=PCA" ) if "CutsGA" in mlist: factory.BookMethod( TMVA.Types.kCuts, "CutsGA", "H:!V:FitMethod=GA:CutRangeMin[0]=-10:CutRangeMax[0]=10:VarProp[1]=FMax:EffSel:Steps=30:Cycles=3:PopSize=400:SC_steps=10:SC_rate=5:SC_factor=0.95" ) if "CutsSA" in mlist: factory.BookMethod( TMVA.Types.kCuts, "CutsSA", "!H:!V:FitMethod=SA:EffSel:MaxCalls=150000:KernelTemp=IncAdaptive:InitialTemp=1e+6:MinTemp=1e-6:Eps=1e-10:UseDefaultScale" ) # Likelihood ("naive Bayes estimator") if "Likelihood" in mlist: factory.BookMethod( TMVA.Types.kLikelihood, "Likelihood", "H:!V:!TransformOutput:PDFInterpol=Spline2:NSmoothSig[0]=20:NSmoothBkg[0]=20:NSmoothBkg[1]=10:NSmooth=1:NAvEvtPerBin=50" ) # Decorrelated likelihood if "LikelihoodD" in mlist: factory.BookMethod( TMVA.Types.kLikelihood, "LikelihoodD", "!H:!V:TransformOutput:PDFInterpol=Spline2:NSmoothSig[0]=20:NSmoothBkg[0]=20:NSmooth=5:NAvEvtPerBin=50:VarTransform=Decorrelate" ) # PCA-transformed likelihood if "LikelihoodPCA" in mlist: factory.BookMethod( TMVA.Types.kLikelihood, "LikelihoodPCA", "!H:!V:!TransformOutput:PDFInterpol=Spline2:NSmoothSig[0]=20:NSmoothBkg[0]=20:NSmooth=5:NAvEvtPerBin=50:VarTransform=PCA" ) # Use a kernel density estimator to approximate the PDFs if "LikelihoodKDE" in mlist: factory.BookMethod( TMVA.Types.kLikelihood, "LikelihoodKDE", "!H:!V:!TransformOutput:PDFInterpol=KDE:KDEtype=Gauss:KDEiter=Adaptive:KDEFineFactor=0.3:KDEborder=None:NAvEvtPerBin=50" ) # Use a variable-dependent mix of splines and kernel density estimator if "LikelihoodMIX" in mlist: factory.BookMethod( TMVA.Types.kLikelihood, "LikelihoodMIX", "!H:!V:!TransformOutput:PDFInterpolSig[0]=KDE:PDFInterpolBkg[0]=KDE:PDFInterpolSig[1]=KDE:PDFInterpolBkg[1]=KDE:PDFInterpolSig[2]=Spline2:PDFInterpolBkg[2]=Spline2:PDFInterpolSig[3]=Spline2:PDFInterpolBkg[3]=Spline2:KDEtype=Gauss:KDEiter=Nonadaptive:KDEborder=None:NAvEvtPerBin=50" ) # Test the multi-dimensional probability density estimator # here are the options strings for the MinMax and RMS methods, respectively: # "!H:!V:VolumeRangeMode=MinMax:DeltaFrac=0.2:KernelEstimator=Gauss:GaussSigma=0.3" ); # "!H:!V:VolumeRangeMode=RMS:DeltaFrac=3:KernelEstimator=Gauss:GaussSigma=0.3" ); if "PDERS" in mlist: factory.BookMethod( TMVA.Types.kPDERS, "PDERS", "!H:!V:NormTree=T:VolumeRangeMode=Adaptive:KernelEstimator=Gauss:GaussSigma=0.3:NEventsMin=400:NEventsMax=600" ) if "PDERSD" in mlist: factory.BookMethod( TMVA.Types.kPDERS, "PDERSD", "!H:!V:VolumeRangeMode=Adaptive:KernelEstimator=Gauss:GaussSigma=0.3:NEventsMin=400:NEventsMax=600:VarTransform=Decorrelate" ) if "PDERSPCA" in mlist: factory.BookMethod( TMVA.Types.kPDERS, "PDERSPCA", "!H:!V:VolumeRangeMode=Adaptive:KernelEstimator=Gauss:GaussSigma=0.3:NEventsMin=400:NEventsMax=600:VarTransform=PCA" ) # Multi-dimensional likelihood estimator using self-adapting phase-space binning if "PDEFoam" in mlist: factory.BookMethod( TMVA.Types.kPDEFoam, "PDEFoam", "!H:!V:SigBgSeparate=F:TailCut=0.001:VolFrac=0.0666:nActiveCells=500:nSampl=2000:nBin=5:Nmin=100:Kernel=None:Compress=T" ) if "PDEFoamBoost" in mlist: factory.BookMethod( TMVA.Types.kPDEFoam, "PDEFoamBoost", "!H:!V:Boost_Num=30:Boost_Transform=linear:SigBgSeparate=F:MaxDepth=4:UseYesNoCell=T:DTLogic=MisClassificationError:FillFoamWithOrigWeights=F:TailCut=0:nActiveCells=500:nBin=20:Nmin=400:Kernel=None:Compress=T" ) # K-Nearest Neighbour classifier (KNN) if "KNN" in mlist: factory.BookMethod( TMVA.Types.kKNN, "KNN", "H:nkNN=20:ScaleFrac=0.8:SigmaFact=1.0:Kernel=Gaus:UseKernel=F:UseWeight=T:!Trim" ) # H-Matrix (chi2-squared) method if "HMatrix" in mlist: factory.BookMethod( TMVA.Types.kHMatrix, "HMatrix", "!H:!V" ) # Linear discriminant (same as Fisher discriminant) if "LD" in mlist: factory.BookMethod( TMVA.Types.kLD, "LD", "H:!V:VarTransform=None:CreateMVAPdfs:PDFInterpolMVAPdf=Spline2:NbinsMVAPdf=50:NsmoothMVAPdf=10" ) # Fisher discriminant (same as LD) if "Fisher" in mlist: factory.BookMethod( TMVA.Types.kFisher, "Fisher", "H:!V:Fisher:CreateMVAPdfs:PDFInterpolMVAPdf=Spline2:NbinsMVAPdf=50:NsmoothMVAPdf=10" ) # Fisher with Gauss-transformed input variables if "FisherG" in mlist: factory.BookMethod( TMVA.Types.kFisher, "FisherG", "H:!V:VarTransform=Gauss" ) # Composite classifier: ensemble (tree) of boosted Fisher classifiers if "BoostedFisher" in mlist: factory.BookMethod( TMVA.Types.kFisher, "BoostedFisher", "H:!V:Boost_Num=20:Boost_Transform=log:Boost_Type=AdaBoost:Boost_AdaBoostBeta=0.2" ) # Function discrimination analysis (FDA) -- test of various fitters - the recommended one is Minuit (or GA or SA) if "FDA_MC" in mlist: factory.BookMethod( TMVA.Types.kFDA, "FDA_MC", "H:!V:Formula=(0)+(1)*x0+(2)*x1+(3)*x2+(4)*x3:ParRanges=(-1,1)(-10,10);(-10,10);(-10,10);(-10,10):FitMethod=MC:SampleSize=100000:Sigma=0.1" ); if "FDA_GA" in mlist: factory.BookMethod( TMVA.Types.kFDA, "FDA_GA", "H:!V:Formula=(0)+(1)*x0+(2)*x1+(3)*x2+(4)*x3:ParRanges=(-1,1)(-10,10);(-10,10);(-10,10);(-10,10):FitMethod=GA:PopSize=300:Cycles=3:Steps=20:Trim=True:SaveBestGen=1" ); if "FDA_SA" in mlist: factory.BookMethod( TMVA.Types.kFDA, "FDA_SA", "H:!V:Formula=(0)+(1)*x0+(2)*x1+(3)*x2+(4)*x3:ParRanges=(-1,1)(-10,10);(-10,10);(-10,10);(-10,10):FitMethod=SA:MaxCalls=15000:KernelTemp=IncAdaptive:InitialTemp=1e+6:MinTemp=1e-6:Eps=1e-10:UseDefaultScale" ); if "FDA_MT" in mlist: factory.BookMethod( TMVA.Types.kFDA, "FDA_MT", "H:!V:Formula=(0)+(1)*x0+(2)*x1+(3)*x2+(4)*x3:ParRanges=(-1,1)(-10,10);(-10,10);(-10,10);(-10,10):FitMethod=MINUIT:ErrorLevel=1:PrintLevel=-1:FitStrategy=2:UseImprove:UseMinos:SetBatch" ); if "FDA_GAMT" in mlist: factory.BookMethod( TMVA.Types.kFDA, "FDA_GAMT", "H:!V:Formula=(0)+(1)*x0+(2)*x1+(3)*x2+(4)*x3:ParRanges=(-1,1)(-10,10);(-10,10);(-10,10);(-10,10):FitMethod=GA:Converger=MINUIT:ErrorLevel=1:PrintLevel=-1:FitStrategy=0:!UseImprove:!UseMinos:SetBatch:Cycles=1:PopSize=5:Steps=5:Trim" ); if "FDA_MCMT" in mlist: factory.BookMethod( TMVA.Types.kFDA, "FDA_MCMT", "H:!V:Formula=(0)+(1)*x0+(2)*x1+(3)*x2+(4)*x3:ParRanges=(-1,1)(-10,10);(-10,10);(-10,10);(-10,10):FitMethod=MC:Converger=MINUIT:ErrorLevel=1:PrintLevel=-1:FitStrategy=0:!UseImprove:!UseMinos:SetBatch:SampleSize=20" ); # TMVA ANN: MLP (recommended ANN) -- all ANNs in TMVA are Multilayer Perceptrons if "MLP" in mlist: factory.BookMethod( TMVA.Types.kMLP, "MLP", "H:!V:NeuronType=tanh:VarTransform=N:NCycles=600:HiddenLayers=N+5:TestRate=5:!UseRegulator" ) if "MLPBFGS" in mlist: factory.BookMethod( TMVA.Types.kMLP, "MLPBFGS", "H:!V:NeuronType=tanh:VarTransform=N:NCycles=600:HiddenLayers=N+5:TestRate=5:TrainingMethod=BFGS:!UseRegulator" ) if "MLPBNN" in mlist: factory.BookMethod( TMVA.Types.kMLP, "MLPBNN", "H:!V:NeuronType=tanh:VarTransform=N:NCycles=600:HiddenLayers=N+5:TestRate=5:TrainingMethod=BFGS:UseRegulator" ) # BFGS training with bayesian regulators # CF(Clermont-Ferrand)ANN if "CFMlpANN" in mlist: factory.BookMethod( TMVA.Types.kCFMlpANN, "CFMlpANN", "!H:!V:NCycles=2000:HiddenLayers=N+1,N" ) # n_cycles:#nodes:#nodes:... # Tmlp(Root)ANN if "TMlpANN" in mlist: factory.BookMethod( TMVA.Types.kTMlpANN, "TMlpANN", "!H:!V:NCycles=200:HiddenLayers=N+1,N:LearningMethod=BFGS:ValidationFraction=0.3" ) # n_cycles:#nodes:#nodes:... # Support Vector Machine if "SVM" in mlist: factory.BookMethod( TMVA.Types.kSVM, "SVM", "Gamma=0.25:Tol=0.001:VarTransform=Norm" ) # Boosted Decision Trees if "BDTG" in mlist: factory.BookMethod( TMVA.Types.kBDT, "BDTG", "!H:!V:NTrees=1000:MinNodeSize=1.5%:BoostType=Grad:Shrinkage=0.10:UseBaggedBoost:BaggedSampleFraction=0.5:nCuts=20:MaxDepth=2" ) if "BDT" in mlist: factory.BookMethod( TMVA.Types.kBDT, "BDT", "!H:!V:NTrees=850:MinNodeSize=2.5%:MaxDepth=3:BoostType=AdaBoost:AdaBoostBeta=0.5:UseBaggedBoost:BaggedSampleFraction=0.5:SeparationType=GiniIndex:nCuts=20" ) if "BDTB" in mlist: factory.BookMethod( TMVA.Types.kBDT, "BDTB", "!H:!V:NTrees=400:BoostType=Bagging:SeparationType=GiniIndex:nCuts=20" ) if "BDTD" in mlist: factory.BookMethod( TMVA.Types.kBDT, "BDTD", "!H:!V:NTrees=400:MinNodeSize=5%:MaxDepth=3:BoostType=AdaBoost:SeparationType=GiniIndex:nCuts=20:VarTransform=Decorrelate" ) # RuleFit -- TMVA implementation of Friedman's method if "RuleFit" in mlist: factory.BookMethod( TMVA.Types.kRuleFit, "RuleFit", "H:!V:RuleFitModule=RFTMVA:Model=ModRuleLinear:MinImp=0.001:RuleMinDist=0.001:NTrees=20:fEventsMin=0.01:fEventsMax=0.5:GDTau=-1.0:GDTauPrec=0.01:GDStep=0.01:GDNSteps=10000:GDErrScale=1.02" ) # -------------------------------------------------------------------------------------------------- # ---- Now you can tell the factory to train, test, and evaluate the MVAs. # Train MVAs factory.TrainAllMethods() # Test MVAs factory.TestAllMethods() # Evaluate MVAs factory.EvaluateAllMethods() # Save the output. outputFile.Close() print "=== wrote root file %s\n" % outfname print "=== TMVAClassification is done!\n" # open the GUI for the result macros gROOT.ProcessLine( "TMVAGui(\"%s\")" % outfname ) # keep the ROOT thread running gApplication.Run() # ---------------------------------------------------------- if __name__ == "__main__": main()