You are here: Foswiki>AtlasSingleTop Web>AnalysisVersion13030>AnalysisTxt13030 (14 Nov 2008, PatRyan)EditAttach

=== TMVAnalysis: use method(s)... === - <CutsGA> === - <Likelihood> === - <LikelihoodPCA> === - <PDERS> === - <KNN> === - <HMatrix> === - <Fisher> === - <FDA_MT> === - <MLP> === - <SVM_Gauss> === - <BDT> === - <BDTD> === - <RuleFit> [1mTMVA -- Toolkit for Multivariate Data Analysis[0m Version 3.9.5, Aug 09, 2008 Copyright (C) 2005-2008 CERN, MPI-K Heidelberg and Victoria U. Home page http://tmva.sourceforge.net All rights reserved, please read http://tmva.sf.net/license.txt TMVAlogon: use "TMVA" style [TMVA style based on "Plain" with modifications defined in tmvaglob.C] ########################################## TMVAout.root --- Factory : You are running ROOT Version: 5.20/00, Jun 24, 2008 --- Factory : --- Factory : _/_/_/_/_/ _| _| _| _| _|_| --- Factory : _/ _|_| _|_| _| _| _| _| --- Factory : _/ _| _| _| _| _| _|_|_|_| --- Factory : _/ _| _| _| _| _| _| --- Factory : _/ _| _| _| _| _| --- Factory : --- Factory : __________TMVA Version 3.9.5, Aug 09, 2008 --- Factory : --- Factory : Preparing trees for training and testing... --- DataSet : Parsing option string: --- DataSet : "NSigTrain=10000000000:NBkgTrain=100000000000::NSigTest=2:NBkgTest=2:SplitMode=Alternate:NormMode=NumEvents:!V" --- DataSet : The following options are set: --- DataSet : - By User: --- DataSet : SplitMode: "Alternate" [Method for selecting training and testing events (default: random)] --- DataSet : NormMode: "NumEvents" [Overall renormalisation of event-by-event weights (NumEvents: average weight of 1 per event, independently for signal and background; EqualNumEvents: average weight of 1 per event for signal, and sum of weights for background equal to sum of weights for signal)] --- DataSet : NSigTrain: "0" [Number of signal training events (default: 0 - all)] --- DataSet : NBkgTrain: "0" [Number of background training events (default: 0 - all)] --- DataSet : NSigTest: "2" [Number of signal testing events (default: 0 - all)] --- DataSet : NBkgTest: "2" [Number of background testing events (default: 0 - all)] --- DataSet : V: "False" [Verbose mode] --- DataSet : - Default: --- DataSet : SplitSeed: "100" [Seed for random event shuffling] --- DataSet : Create training and testing trees: looping over signal events ... --- DataSet : Create training and testing trees: looping over background events ... --- DataSet : Prepare training and test samples: --- DataSet : Signal tree - number of events : 518 --- DataSet : Background tree - number of events : 825 --- DataSet : Preselection: --- DataSet : No preselection cuts applied on signal or background --- DataSet : Weight renormalisation mode: "NumEvents": renormalise signal and background --- DataSet : weights independently so that Sum[i=1..N_j]{w_i} = N_j, j=signal, background --- DataSet : (note that N_j is the sum of training and test events) --- DataSet : Event weights scaling factor: --- DataSet : signal : 5.7503 --- DataSet : background : 1.76908 --- DataSet : Pick alternating training and test events from input tree for signal --- DataSet : Pick alternating training and test events from input tree for background --- DataSet : Create training tree --- DataSet : Create testing tree --- DataSet : Collected: --- DataSet : - Training signal entries : 516 --- DataSet : - Training background entries : 823 --- DataSet : - Testing signal entries : 2 --- DataSet : - Testing background entries : 2 --- DataSet : Compute correlation matrices on tree: TrainingTree --- DataSet : --- DataSet : Correlation matrix (signal): --- DataSet : --------------------------------------------------------------- --- DataSet : HT Jet1Pt DeltaRJet1Jet2 WTransverseMass --- DataSet : HT: +1.000 +0.932 +0.125 +0.181 --- DataSet : Jet1Pt: +0.932 +1.000 +0.191 +0.038 --- DataSet : DeltaRJet1Jet2: +0.125 +0.191 +1.000 -0.226 --- DataSet : WTransverseMass: +0.181 +0.038 -0.226 +1.000 --- DataSet : --------------------------------------------------------------- --- DataSet : --- DataSet : Correlation matrix (background): --- DataSet : --------------------------------------------------------------- --- DataSet : HT Jet1Pt DeltaRJet1Jet2 WTransverseMass --- DataSet : HT: +1.000 +0.927 +0.141 +0.018 --- DataSet : Jet1Pt: +0.927 +1.000 +0.214 -0.069 --- DataSet : DeltaRJet1Jet2: +0.141 +0.214 +1.000 -0.257 --- DataSet : WTransverseMass: +0.018 -0.069 -0.257 +1.000 --- DataSet : --------------------------------------------------------------- --- DataSet : --- DataSet : New variable Transformation NoTransform requested and created. --- TransBase : Create scatter and profile plots in target-file directory: --- TransBase : TMVAout.root:/InputVariables_NoTransform/CorrelationPlots --- TransBase : Ranking input variables... --- NoTransform : Ranking result (top variable is best ranked) --- NoTransform : ---------------------------------------------------------------- --- NoTransform : Rank : Variable : Separation --- NoTransform : ---------------------------------------------------------------- --- NoTransform : 1 : HT : 2.563e-01 --- NoTransform : 2 : Jet1Pt : 2.270e-01 --- NoTransform : 3 : WTransverseMass : 7.929e-02 --- NoTransform : 4 : DeltaRJet1Jet2 : 6.387e-02 --- NoTransform : ---------------------------------------------------------------- --- Cuts : Parsing option string: --- Cuts : "H:!V:FitMethod=GA:EffSel:Steps=30:Cycles=3:PopSize=100:SC_steps=10:SC_rate=5:SC_factor=0.95:VarProp=FSmart" --- Cuts : The following options are set: --- Cuts : - By User: --- Cuts : V: "False" [Verbose mode] --- Cuts : H: "True" [Print classifier-specific help message] --- Cuts : FitMethod: "GA" [Minimization Method (GA, SA, and MC are the primary methods to be used; the others have been introduced for testing purposes and are depreciated)] --- Cuts : EffMethod: "EffSel" [Selection Method] --- Cuts : - Default: --- Cuts : D: "False" [Use-decorrelated-variables flag (depreciated)] --- Cuts : Normalise: "False" [Normalise input variables] --- Cuts : VarTransform: "None" [Variable transformation method] --- Cuts : VarTransformType: "Signal" [Use signal or background events to derive for variable transformation (the transformation is applied on both types of, course)] --- Cuts : NbinsMVAPdf: "60" [Number of bins used for the PDFs of classifier outputs] --- Cuts : NsmoothMVAPdf: "2" [Number of smoothing iterations for classifier PDFs] --- Cuts : VerboseLevel: "Default" [Verbosity level] --- Cuts : CreateMVAPdfs: "False" [Create PDFs for classifier outputs (signal and background)] --- Cuts : TxtWeightFilesOnly: "True" [If True: write all training results (weights) as text files (False: some are written in ROOT format)] --- Cuts : CutRangeMin[0]: "-1" [Minimum of allowed cut range (set per variable)] --- Cuts : CutRangeMin[1]: "-1" --- Cuts : CutRangeMin[2]: "-1" --- Cuts : CutRangeMin[3]: "-1" --- Cuts : CutRangeMax[0]: "-1" [Maximum of allowed cut range (set per variable)] --- Cuts : CutRangeMax[1]: "-1" --- Cuts : CutRangeMax[2]: "-1" --- Cuts : CutRangeMax[3]: "-1" --- Cuts : VarProp[0]: "FSmart" [Categorisation of cuts] --- Cuts : VarProp[1]: "FSmart" --- Cuts : VarProp[2]: "FSmart" --- Cuts : VarProp[3]: "FSmart" --- Cuts : Use optimization method: 'Genetic Algorithm' --- Cuts : Use efficiency computation method: 'Event Selection' --- Cuts : Use "FSmart" cuts for variable: 'HT' --- Cuts : Use "FSmart" cuts for variable: 'Jet1Pt' --- Cuts : Use "FSmart" cuts for variable: 'DeltaRJet1Jet2' --- Cuts : Use "FSmart" cuts for variable: 'WTransverseMass' --- Factory : Booking method: CutsGA --- Likelihood : Parsing option string: --- Likelihood : "!H:!V:!TransformOutput:PDFInterpol=Spline2:NSmoothSig[0]=10:NSmoothBkg[0]=10:NSmoothBkg[1]=10:NSmooth=10:NAvEvtPerBin=50" --- Likelihood : The following options are set: --- Likelihood : - By User: --- Likelihood : V: "False" [Verbose mode] --- Likelihood : H: "False" [Print classifier-specific help message] --- Likelihood : NSmooth: "10" [Number of smoothing iterations for the input histograms] --- Likelihood : NAvEvtPerBin: "50" [Average number of events per PDF bin] --- Likelihood : TransformOutput: "False" [Transform likelihood output by inverse sigmoid function] --- Likelihood : - Default: --- Likelihood : D: "False" [Use-decorrelated-variables flag (depreciated)] --- Likelihood : Normalise: "False" [Normalise input variables] --- Likelihood : VarTransform: "None" [Variable transformation method] --- Likelihood : VarTransformType: "Signal" [Use signal or background events to derive for variable transformation (the transformation is applied on both types of, course)] --- Likelihood : NbinsMVAPdf: "60" [Number of bins used for the PDFs of classifier outputs] --- Likelihood : NsmoothMVAPdf: "2" [Number of smoothing iterations for classifier PDFs] --- Likelihood : VerboseLevel: "Default" [Verbosity level] --- Likelihood : CreateMVAPdfs: "False" [Create PDFs for classifier outputs (signal and background)] --- Likelihood : TxtWeightFilesOnly: "True" [If True: write all training results (weights) as text files (False: some are written in ROOT format)] --- Likelihood : NSmoothSig[0]: "10" [Number of smoothing iterations for the input histograms] --- Likelihood : NSmoothSig[1]: "-1" --- Likelihood : NSmoothSig[2]: "-1" --- Likelihood : NSmoothSig[3]: "-1" --- Likelihood : NSmoothBkg[0]: "10" [Number of smoothing iterations for the input histograms] --- Likelihood : NSmoothBkg[1]: "10" --- Likelihood : NSmoothBkg[2]: "-1" --- Likelihood : NSmoothBkg[3]: "-1" --- Likelihood : NAvEvtPerBinSig[0]: "-1" [Average num of events per PDF bin and variable (signal)] --- Likelihood : NAvEvtPerBinSig[1]: "-1" --- Likelihood : NAvEvtPerBinSig[2]: "-1" --- Likelihood : NAvEvtPerBinSig[3]: "-1" --- Likelihood : NAvEvtPerBinBkg[0]: "-1" [Average num of events per PDF bin and variable (background)] --- Likelihood : NAvEvtPerBinBkg[1]: "-1" --- Likelihood : NAvEvtPerBinBkg[2]: "-1" --- Likelihood : NAvEvtPerBinBkg[3]: "-1" --- Likelihood : PDFInterpol[0]: "Spline2" [Method of interpolating reference histograms (e.g. Spline2 or KDE)] --- Likelihood : PDFInterpol[1]: "Spline2" --- Likelihood : PDFInterpol[2]: "Spline2" --- Likelihood : PDFInterpol[3]: "Spline2" --- Likelihood : KDEtype: "Gauss" [KDE kernel type (1=Gauss)] --- Likelihood : KDEiter: "Nonadaptive" [Number of iterations (1=non-adaptive, 2=adaptive)] --- Likelihood : KDEFineFactor: "1" [Fine tuning factor for Adaptive KDE: Factor to multyply the width of the kernel] --- Likelihood : KDEborder: "None" [Border effects treatment (1=no treatment , 2=kernel renormalization, 3=sample mirroring)] --- Factory : Booking method: Likelihood --- Likelihood : Parsing option string: --- Likelihood : "!H:!V:!TransformOutput:PDFInterpol=Spline2:NSmoothSig[0]=10:NSmoothBkg[0]=10:NSmooth=5:NAvEvtPerBin=50:VarTransform=PCA" --- Likelihood : The following options are set: --- Likelihood : - By User: --- Likelihood : VarTransform: "PCA" [Variable transformation method] --- Likelihood : V: "False" [Verbose mode] --- Likelihood : H: "False" [Print classifier-specific help message] --- Likelihood : NSmooth: "5" [Number of smoothing iterations for the input histograms] --- Likelihood : NAvEvtPerBin: "50" [Average number of events per PDF bin] --- Likelihood : TransformOutput: "False" [Transform likelihood output by inverse sigmoid function] --- Likelihood : - Default: --- Likelihood : D: "False" [Use-decorrelated-variables flag (depreciated)] --- Likelihood : Normalise: "False" [Normalise input variables] --- Likelihood : VarTransformType: "Signal" [Use signal or background events to derive for variable transformation (the transformation is applied on both types of, course)] --- Likelihood : NbinsMVAPdf: "60" [Number of bins used for the PDFs of classifier outputs] --- Likelihood : NsmoothMVAPdf: "2" [Number of smoothing iterations for classifier PDFs] --- Likelihood : VerboseLevel: "Default" [Verbosity level] --- Likelihood : CreateMVAPdfs: "False" [Create PDFs for classifier outputs (signal and background)] --- Likelihood : TxtWeightFilesOnly: "True" [If True: write all training results (weights) as text files (False: some are written in ROOT format)] --- Likelihood : NSmoothSig[0]: "10" [Number of smoothing iterations for the input histograms] --- Likelihood : NSmoothSig[1]: "-1" --- Likelihood : NSmoothSig[2]: "-1" --- Likelihood : NSmoothSig[3]: "-1" --- Likelihood : NSmoothBkg[0]: "10" [Number of smoothing iterations for the input histograms] --- Likelihood : NSmoothBkg[1]: "-1" --- Likelihood : NSmoothBkg[2]: "-1" --- Likelihood : NSmoothBkg[3]: "-1" --- Likelihood : NAvEvtPerBinSig[0]: "-1" [Average num of events per PDF bin and variable (signal)] --- Likelihood : NAvEvtPerBinSig[1]: "-1" --- Likelihood : NAvEvtPerBinSig[2]: "-1" --- Likelihood : NAvEvtPerBinSig[3]: "-1" --- Likelihood : NAvEvtPerBinBkg[0]: "-1" [Average num of events per PDF bin and variable (background)] --- Likelihood : NAvEvtPerBinBkg[1]: "-1" --- Likelihood : NAvEvtPerBinBkg[2]: "-1" --- Likelihood : NAvEvtPerBinBkg[3]: "-1" --- Likelihood : PDFInterpol[0]: "Spline2" [Method of interpolating reference histograms (e.g. Spline2 or KDE)] --- Likelihood : PDFInterpol[1]: "Spline2" --- Likelihood : PDFInterpol[2]: "Spline2" --- Likelihood : PDFInterpol[3]: "Spline2" --- Likelihood : KDEtype: "Gauss" [KDE kernel type (1=Gauss)] --- Likelihood : KDEiter: "Nonadaptive" [Number of iterations (1=non-adaptive, 2=adaptive)] --- Likelihood : KDEFineFactor: "1" [Fine tuning factor for Adaptive KDE: Factor to multyply the width of the kernel] --- Likelihood : KDEborder: "None" [Border effects treatment (1=no treatment , 2=kernel renormalization, 3=sample mirroring)] --- DataSet : New variable Transformation PCATransform requested and created. --- TransBase : Create scatter and profile plots in target-file directory: --- TransBase : TMVAout.root:/InputVariables_PCATransform/CorrelationPlots --- TransBase : Ranking input variables... --- PCATransform : Ranking result (top variable is best ranked) --- PCATransform : ---------------------------------------------------------------- --- PCATransform : Rank : Variable : Separation --- PCATransform : ---------------------------------------------------------------- --- PCATransform : 1 : HT : 2.483e-01 --- PCATransform : 2 : DeltaRJet1Jet2 : 9.027e-02 --- PCATransform : 3 : Jet1Pt : 8.891e-02 --- PCATransform : 4 : WTransverseMass : 5.416e-02 --- PCATransform : ---------------------------------------------------------------- --- Likelihood : Use principal component transformation --- Factory : Booking method: LikelihoodPCA --- PDERS : Parsing option string: --- PDERS : "!H:!V:NormTree=T:VolumeRangeMode=Adaptive:KernelEstimator=Gauss:GaussSigma=0.3:NEventsMin=400:NEventsMax=600" --- PDERS : The following options are set: --- PDERS : - By User: --- PDERS : V: "False" [Verbose mode] --- PDERS : H: "False" [Print classifier-specific help message] --- PDERS : VolumeRangeMode: "Adaptive" [Method to determine volume size] --- PDERS : KernelEstimator: "Gauss" [Kernel estimation function] --- PDERS : NEventsMin: "400" [nEventsMin for adaptive volume range] --- PDERS : NEventsMax: "600" [nEventsMax for adaptive volume range] --- PDERS : GaussSigma: "0.3" [Width (wrt volume size) of Gaussian kernel estimator] --- PDERS : NormTree: "True" [Normalize binary search tree] --- PDERS : - Default: --- PDERS : D: "False" [Use-decorrelated-variables flag (depreciated)] --- PDERS : Normalise: "False" [Normalise input variables] --- PDERS : VarTransform: "None" [Variable transformation method] --- PDERS : VarTransformType: "Signal" [Use signal or background events to derive for variable transformation (the transformation is applied on both types of, course)] --- PDERS : NbinsMVAPdf: "60" [Number of bins used for the PDFs of classifier outputs] --- PDERS : NsmoothMVAPdf: "2" [Number of smoothing iterations for classifier PDFs] --- PDERS : VerboseLevel: "Default" [Verbosity level] --- PDERS : CreateMVAPdfs: "False" [Create PDFs for classifier outputs (signal and background)] --- PDERS : TxtWeightFilesOnly: "True" [If True: write all training results (weights) as text files (False: some are written in ROOT format)] --- PDERS : DeltaFrac: "3" [nEventsMin/Max for minmax and rms volume range] --- PDERS : MaxVIterations: "150" [MaxVIterations for adaptive volume range] --- PDERS : InitialScale: "0.99" [InitialScale for adaptive volume range] --- Factory : Booking method: PDERS --- KNN : Parsing option string: --- KNN : "nkNN=400:TreeOptDepth=6:ScaleFrac=0.8:!UseKernel:!Trim" --- KNN : The following options are set: --- KNN : - By User: --- KNN : nkNN: "400" [Number of k-nearest neighbors] --- KNN : TreeOptDepth: "6" [Binary tree optimisation depth] --- KNN : ScaleFrac: "0.8" [Fraction of events used for scaling] --- KNN : UseKernel: "False" [Use polynomial kernel weight] --- KNN : Trim: "False" [Use equal number of signal and background events] --- KNN : - Default: --- KNN : D: "False" [Use-decorrelated-variables flag (depreciated)] --- KNN : Normalise: "False" [Normalise input variables] --- KNN : VarTransform: "None" [Variable transformation method] --- KNN : VarTransformType: "Signal" [Use signal or background events to derive for variable transformation (the transformation is applied on both types of, course)] --- KNN : NbinsMVAPdf: "60" [Number of bins used for the PDFs of classifier outputs] --- KNN : NsmoothMVAPdf: "2" [Number of smoothing iterations for classifier PDFs] --- KNN : V: "False" [Verbose mode] --- KNN : VerboseLevel: "Default" [Verbosity level] --- KNN : H: "False" [Print classifier-specific help message] --- KNN : CreateMVAPdfs: "False" [Create PDFs for classifier outputs (signal and background)] --- KNN : TxtWeightFilesOnly: "True" [If True: write all training results (weights) as text files (False: some are written in ROOT format)] --- Factory : Booking method: KNN --- HMatrix : Parsing option string: --- HMatrix : "!H:!V" --- HMatrix : The following options are set: --- HMatrix : - By User: --- HMatrix : V: "False" [Verbose mode] --- HMatrix : H: "False" [Print classifier-specific help message] --- HMatrix : - Default: --- HMatrix : D: "False" [Use-decorrelated-variables flag (depreciated)] --- HMatrix : Normalise: "True" [Normalise input variables] --- HMatrix : VarTransform: "None" [Variable transformation method] --- HMatrix : VarTransformType: "Signal" [Use signal or background events to derive for variable transformation (the transformation is applied on both types of, course)] --- HMatrix : NbinsMVAPdf: "60" [Number of bins used for the PDFs of classifier outputs] --- HMatrix : NsmoothMVAPdf: "2" [Number of smoothing iterations for classifier PDFs] --- HMatrix : VerboseLevel: "Default" [Verbosity level] --- HMatrix : CreateMVAPdfs: "False" [Create PDFs for classifier outputs (signal and background)] --- HMatrix : TxtWeightFilesOnly: "True" [If True: write all training results (weights) as text files (False: some are written in ROOT format)] --- Factory : Booking method: HMatrix --- Fisher : Parsing option string: --- Fisher : "H:!V:!Normalise:CreateMVAPdfs:Fisher:NbinsMVAPdf=50:NsmoothMVAPdf=1" --- Fisher : The following options are set: --- Fisher : - By User: --- Fisher : Normalise: "False" [Normalise input variables] --- Fisher : NbinsMVAPdf: "50" [Number of bins used for the PDFs of classifier outputs] --- Fisher : NsmoothMVAPdf: "1" [Number of smoothing iterations for classifier PDFs] --- Fisher : V: "False" [Verbose mode] --- Fisher : H: "True" [Print classifier-specific help message] --- Fisher : CreateMVAPdfs: "True" [Create PDFs for classifier outputs (signal and background)] --- Fisher : Method: "Fisher" [Discrimination method] --- Fisher : - Default: --- Fisher : D: "False" [Use-decorrelated-variables flag (depreciated)] --- Fisher : VarTransform: "None" [Variable transformation method] --- Fisher : VarTransformType: "Signal" [Use signal or background events to derive for variable transformation (the transformation is applied on both types of, course)] --- Fisher : VerboseLevel: "Default" [Verbosity level] --- Fisher : TxtWeightFilesOnly: "True" [If True: write all training results (weights) as text files (False: some are written in ROOT format)] --- Factory : Booking method: Fisher --- FDA : Parsing option string: --- FDA : "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" --- FDA : The following options are set: --- FDA : - By User: --- FDA : V: "False" [Verbose mode] --- FDA : H: "True" [Print classifier-specific help message] --- FDA : Formula: "(0)+(1)*x0+(2)*x1+(3)*x2+(4)*x3" [The discrimination formula] --- FDA : ParRanges: "(-1,1);(-10,10);(-10,10);(-10,10);(-10,10)" [Parameter ranges] --- FDA : FitMethod: "MINUIT" [Optimisation Method] --- FDA : - Default: --- FDA : D: "False" [Use-decorrelated-variables flag (depreciated)] --- FDA : Normalise: "False" [Normalise input variables] --- FDA : VarTransform: "None" [Variable transformation method] --- FDA : VarTransformType: "Signal" [Use signal or background events to derive for variable transformation (the transformation is applied on both types of, course)] --- FDA : NbinsMVAPdf: "60" [Number of bins used for the PDFs of classifier outputs] --- FDA : NsmoothMVAPdf: "2" [Number of smoothing iterations for classifier PDFs] --- FDA : VerboseLevel: "Default" [Verbosity level] --- FDA : CreateMVAPdfs: "False" [Create PDFs for classifier outputs (signal and background)] --- FDA : TxtWeightFilesOnly: "True" [If True: write all training results (weights) as text files (False: some are written in ROOT format)] --- FDA : Converger: "None" [FitMethod uses Converger to improve result] --- FDA : User-defined formula string : "(0)+(1)*x0+(2)*x1+(3)*x2+(4)*x3" --- FDA : TFormula-compatible formula string: "[0]+[1]*[5]+[2]*[6]+[3]*[7]+[4]*[8]" --- FDA : Creating and compiling formula --- FDA_Fitter_M...: Parsing option string: --- FDA_Fitter_M...: "!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" --- FDA_Fitter_M...: The following options are set: --- FDA_Fitter_M...: - By User: --- FDA_Fitter_M...: ErrorLevel: "1" [TMinuit: error level: 0.5=logL fit, 1=chi-squared fit] --- FDA_Fitter_M...: PrintLevel: "-1" [TMinuit: output level: -1=least, 0, +1=all garbage] --- FDA_Fitter_M...: FitStrategy: "2" [TMinuit: fit strategy: 2=best] --- FDA_Fitter_M...: UseImprove: "True" [TMinuit: use IMPROVE] --- FDA_Fitter_M...: UseMinos: "True" [TMinuit: use MINOS] --- FDA_Fitter_M...: SetBatch: "True" [TMinuit: use batch mode] --- FDA_Fitter_M...: - Default: --- FDA_Fitter_M...: PrintWarnings: "False" [TMinuit: suppress warnings] --- FDA_Fitter_M...: MaxCalls: "1000" [TMinuit: approximate maximum number of function calls] --- FDA_Fitter_M...: Tolerance: "0.1" [TMinuit: tolerance to the function value at the minimum] --- Factory : Booking method: FDA_MT --- MLP : Parsing option string: --- MLP : "H:!V:!Normalise:NeuronType=tanh:NCycles=200:HiddenLayers=N+1,N:TestRate=5" --- MLP : The following options are set: --- MLP : - By User: --- MLP : Normalise: "False" [Normalise input variables] --- MLP : V: "False" [Verbose mode] --- MLP : H: "True" [Print classifier-specific help message] --- MLP : NCycles: "200" [Number of training cycles] --- MLP : HiddenLayers: "N+1,N" [Specification of hidden layer architecture (N stands for number of variables; any integers may also be used)] --- MLP : NeuronType: "tanh" [Neuron activation function type] --- MLP : TestRate: "5" [Test for overtraining performed at each #th epochs] --- MLP : - Default: --- MLP : D: "False" [Use-decorrelated-variables flag (depreciated)] --- MLP : VarTransform: "None" [Variable transformation method] --- MLP : VarTransformType: "Signal" [Use signal or background events to derive for variable transformation (the transformation is applied on both types of, course)] --- MLP : NbinsMVAPdf: "60" [Number of bins used for the PDFs of classifier outputs] --- MLP : NsmoothMVAPdf: "2" [Number of smoothing iterations for classifier PDFs] --- MLP : VerboseLevel: "Default" [Verbosity level] --- MLP : CreateMVAPdfs: "False" [Create PDFs for classifier outputs (signal and background)] --- MLP : TxtWeightFilesOnly: "True" [If True: write all training results (weights) as text files (False: some are written in ROOT format)] --- MLP : NeuronInputType: "sum" [Neuron input function type] --- MLP : TrainingMethod: "BP" [Train with Back-Propagation (BP - default) or Genetic Algorithm (GA - slower and worse)] --- MLP : LearningRate: "0.02" [ANN learning rate parameter] --- MLP : DecayRate: "0.01" [Decay rate for learning parameter] --- MLP : BPMode: "sequential" [Back-propagation learning mode: sequential or batch] --- MLP : BatchSize: "-1" [Batch size: number of events/batch, only set if in Batch Mode, -1 for BatchSize=number_of_events] --- MLP : Building Network --- MLP : Initializing weights --- Factory : Booking method: MLP --- SVM : Parsing option string: --- SVM : "Sigma=2:C=1:Tol=0.001:Kernel=Gauss" --- SVM : The following options are set: --- SVM : - By User: --- SVM : C: "1" [C parameter] --- SVM : Tol: "0.001" [Tolerance parameter] --- SVM : Sigma: "2" [Kernel parameter: sigma] --- SVM : Kernel: "Gauss" [Uses kernel function] --- SVM : - Default: --- SVM : D: "False" [Use-decorrelated-variables flag (depreciated)] --- SVM : Normalise: "True" [Normalise input variables] --- SVM : VarTransform: "None" [Variable transformation method] --- SVM : VarTransformType: "Signal" [Use signal or background events to derive for variable transformation (the transformation is applied on both types of, course)] --- SVM : NbinsMVAPdf: "60" [Number of bins used for the PDFs of classifier outputs] --- SVM : NsmoothMVAPdf: "2" [Number of smoothing iterations for classifier PDFs] --- SVM : V: "False" [Verbose mode] --- SVM : VerboseLevel: "Default" [Verbosity level] --- SVM : H: "False" [Print classifier-specific help message] --- SVM : CreateMVAPdfs: "False" [Create PDFs for classifier outputs (signal and background)] --- SVM : TxtWeightFilesOnly: "True" [If True: write all training results (weights) as text files (False: some are written in ROOT format)] --- SVM : MaxIter: "1000" [Maximum number of training loops] --- SVM : Order: "3" [Polynomial Kernel parameter: polynomial order] --- SVM : Theta: "1" [Sigmoid Kernel parameter: theta] --- SVM : Kappa: "1" [Sigmoid Kernel parameter: kappa] --- Factory : Booking method: SVM_Gauss --- BDT : Parsing option string: --- BDT : "!H:!V:NTrees=400:BoostType=AdaBoost:SeparationType=GiniIndex:nCuts=20:PruneMethod=CostComplexity:PruneStrength=1.5" --- BDT : The following options are set: --- BDT : - By User: --- BDT : V: "False" [Verbose mode] --- BDT : H: "False" [Print classifier-specific help message] --- BDT : NTrees: "400" [Number of trees in the forest] --- BDT : BoostType: "AdaBoost" [Boosting type for the trees in the forest] --- BDT : SeparationType: "GiniIndex" [Separation criterion for node splitting] --- BDT : nCuts: "20" [Number of steps during node cut optimisation] --- BDT : PruneStrength: "1.5" [Pruning strength] --- BDT : PruneMethod: "CostComplexity" [Method used for pruning (removal) of statistically insignificant branches] --- BDT : - Default: --- BDT : D: "False" [Use-decorrelated-variables flag (depreciated)] --- BDT : Normalise: "False" [Normalise input variables] --- BDT : VarTransform: "None" [Variable transformation method] --- BDT : VarTransformType: "Signal" [Use signal or background events to derive for variable transformation (the transformation is applied on both types of, course)] --- BDT : NbinsMVAPdf: "60" [Number of bins used for the PDFs of classifier outputs] --- BDT : NsmoothMVAPdf: "2" [Number of smoothing iterations for classifier PDFs] --- BDT : VerboseLevel: "Default" [Verbosity level] --- BDT : CreateMVAPdfs: "False" [Create PDFs for classifier outputs (signal and background)] --- BDT : TxtWeightFilesOnly: "True" [If True: write all training results (weights) as text files (False: some are written in ROOT format)] --- BDT : AdaBoostBeta: "1" [Parameter for AdaBoost algorithm] --- BDT : UseRandomisedTrees: "False" [Choose at each node splitting a random set of variables] --- BDT : UseNvars: "4" [Number of variables used if randomised Tree option is chosen] --- BDT : UseWeightedTrees: "True" [Use weighted trees or simple average in classification from the forest] --- BDT : UseYesNoLeaf: "True" [Use Sig or Bkg categories, or the purity=S/(S+B) as classification of the leaf node] --- BDT : NodePurityLimit: "0.5" [In boosting/pruning, nodes with purity > NodePurityLimit are signal; background otherwise.] --- BDT : nEventsMin: "20" [Minimum number of events required in a leaf node (default: max(20, N_train/(Nvar^2)/10) ) ] --- BDT : PruneBeforeBoost: "False" [Flag to prune the tree before applying boosting algorithm] --- BDT : NoNegWeightsInTraining: "False" [Ignore negative event weights in the training process] --- BDT : Events with negative event weights are ignored during the BDT training (option NoNegWeightsInTraining=0 --- BDT : <InitEventSample> Internally I use 1339 for Training and 0 for Validation --- Factory : Booking method: BDT --- BDT : Parsing option string: --- BDT : "!H:!V:NTrees=400:BoostType=AdaBoost:SeparationType=GiniIndex:nCuts=20:PruneMethod=CostComplexity:PruneStrength=1.5:VarTransform=Decorrelate" --- BDT : The following options are set: --- BDT : - By User: --- BDT : VarTransform: "Decorrelate" [Variable transformation method] --- BDT : V: "False" [Verbose mode] --- BDT : H: "False" [Print classifier-specific help message] --- BDT : NTrees: "400" [Number of trees in the forest] --- BDT : BoostType: "AdaBoost" [Boosting type for the trees in the forest] --- BDT : SeparationType: "GiniIndex" [Separation criterion for node splitting] --- BDT : nCuts: "20" [Number of steps during node cut optimisation] --- BDT : PruneStrength: "1.5" [Pruning strength] --- BDT : PruneMethod: "CostComplexity" [Method used for pruning (removal) of statistically insignificant branches] --- BDT : - Default: --- BDT : D: "False" [Use-decorrelated-variables flag (depreciated)] --- BDT : Normalise: "False" [Normalise input variables] --- BDT : VarTransformType: "Signal" [Use signal or background events to derive for variable transformation (the transformation is applied on both types of, course)] --- BDT : NbinsMVAPdf: "60" [Number of bins used for the PDFs of classifier outputs] --- BDT : NsmoothMVAPdf: "2" [Number of smoothing iterations for classifier PDFs] --- BDT : VerboseLevel: "Default" [Verbosity level] --- BDT : CreateMVAPdfs: "False" [Create PDFs for classifier outputs (signal and background)] --- BDT : TxtWeightFilesOnly: "True" [If True: write all training results (weights) as text files (False: some are written in ROOT format)] --- BDT : AdaBoostBeta: "1" [Parameter for AdaBoost algorithm] --- BDT : UseRandomisedTrees: "False" [Choose at each node splitting a random set of variables] --- BDT : UseNvars: "4" [Number of variables used if randomised Tree option is chosen] --- BDT : UseWeightedTrees: "True" [Use weighted trees or simple average in classification from the forest] --- BDT : UseYesNoLeaf: "True" [Use Sig or Bkg categories, or the purity=S/(S+B) as classification of the leaf node] --- BDT : NodePurityLimit: "0.5" [In boosting/pruning, nodes with purity > NodePurityLimit are signal; background otherwise.] --- BDT : nEventsMin: "20" [Minimum number of events required in a leaf node (default: max(20, N_train/(Nvar^2)/10) ) ] --- BDT : PruneBeforeBoost: "False" [Flag to prune the tree before applying boosting algorithm] --- BDT : NoNegWeightsInTraining: "False" [Ignore negative event weights in the training process] --- DataSet : New variable Transformation DecorrTransform requested and created. --- TransBase : Create scatter and profile plots in target-file directory: --- TransBase : TMVAout.root:/InputVariables_DecorrTransform/CorrelationPlots --- TransBase : Ranking input variables... --- DecorrTransform: Ranking result (top variable is best ranked) --- DecorrTransform: ---------------------------------------------------------------- --- DecorrTransform: Rank : Variable : Separation --- DecorrTransform: ---------------------------------------------------------------- --- DecorrTransform: 1 : HT : 2.083e-01 --- DecorrTransform: 2 : Jet1Pt : 8.879e-02 --- DecorrTransform: 3 : WTransverseMass : 8.553e-02 --- DecorrTransform: 4 : DeltaRJet1Jet2 : 4.945e-02 --- DecorrTransform: ---------------------------------------------------------------- --- BDT : Events with negative event weights are ignored during the BDT training (option NoNegWeightsInTraining=0 --- BDT : <InitEventSample> Internally I use 1339 for Training and 0 for Validation --- Factory : Booking method: BDTD --- RuleFit : Parsing option string: --- 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" --- RuleFit : The following options are set: --- RuleFit : - By User: --- RuleFit : V: "False" [Verbose mode] --- RuleFit : H: "True" [Print classifier-specific help message] --- RuleFit : GDTau: "-1" [Gradient-directed path: default fit cut-off] --- RuleFit : GDTauPrec: "0.01" [Gradient-directed path: precision of tau] --- RuleFit : GDStep: "0.01" [Gradient-directed path: step size] --- RuleFit : GDNSteps: "10000" [Gradient-directed path: number of steps] --- RuleFit : GDErrScale: "1.02" [Stop scan when error>scale*errmin] --- RuleFit : fEventsMin: "0.01" [Minimum fraction of events in a splittable node] --- RuleFit : fEventsMax: "0.5" [Maximum fraction of events in a splittable node] --- RuleFit : nTrees: "20" [Number of trees in forest.] --- RuleFit : RuleMinDist: "0.001" [Minimum distance between rules] --- RuleFit : MinImp: "0.001" [Minimum rule importance accepted] --- RuleFit : Model: "ModRuleLinear" [Model to be used] --- RuleFit : RuleFitModule: "RFTMVA" [Which RuleFit module to use] --- RuleFit : - Default: --- RuleFit : D: "False" [Use-decorrelated-variables flag (depreciated)] --- RuleFit : Normalise: "False" [Normalise input variables] --- RuleFit : VarTransform: "None" [Variable transformation method] --- RuleFit : VarTransformType: "Signal" [Use signal or background events to derive for variable transformation (the transformation is applied on both types of, course)] --- RuleFit : NbinsMVAPdf: "60" [Number of bins used for the PDFs of classifier outputs] --- RuleFit : NsmoothMVAPdf: "2" [Number of smoothing iterations for classifier PDFs] --- RuleFit : VerboseLevel: "Default" [Verbosity level] --- RuleFit : CreateMVAPdfs: "False" [Create PDFs for classifier outputs (signal and background)] --- RuleFit : TxtWeightFilesOnly: "True" [If True: write all training results (weights) as text files (False: some are written in ROOT format)] --- RuleFit : GDPathEveFrac: "0.5" [Fraction of events used for the path search] --- RuleFit : GDValidEveFrac: "0.5" [Fraction of events used for the validation] --- RuleFit : ForestType: "AdaBoost" [Method to use for forest generation] --- RuleFit : RFWorkDir: "./rulefit" [Friedmans RuleFit module: working dir] --- RuleFit : RFNrules: "2000" [Friedmans RuleFit module: maximum number of rules] --- RuleFit : RFNendnodes: "4" [Friedmans RuleFit module: average number of end nodes] --- Factory : Booking method: RuleFit --- Factory : Training all methods... --- Factory : Train method: CutsGA --- Cuts : Option for variable: HT: 'ForceSmart' (#: 3) --- Cuts : Option for variable: Jet1Pt: 'ForceSmart' (#: 3) --- Cuts : Option for variable: DeltaRJet1Jet2: 'ForceSmart' (#: 3) --- Cuts : Option for variable: WTransverseMass: 'ForceSmart' (#: 3) --- Cuts : --- Cuts : [1m================================================================[0m --- Cuts : [1mH e l p f o r c l a s s i f i e r [ Cuts ] :[0m --- Cuts : --- Cuts : [1m--- Short description:[0m --- Cuts : --- Cuts : The optimisation of rectangular cuts performed by TMVA maximises --- Cuts : the background rejection at given signal efficiency, and scans --- Cuts : over the full range of the latter quantity. Three optimisation --- Cuts : methods are optional: Monte Carlo sampling (MC), a Genetics --- Cuts : Algorithm (GA), and Simulated Annealing (SA). GA and SA are --- Cuts : expected to perform best. --- Cuts : --- Cuts : The difficulty to find the optimal cuts strongly increases with --- Cuts : the dimensionality (number of input variables) of the problem. --- Cuts : This behavior is due to the non-uniqueness of the solution space. --- Cuts : --- Cuts : [1m--- Performance optimisation:[0m --- Cuts : --- Cuts : If the dimensionality exceeds, say, 4 input variables, it is --- Cuts : advisable to scrutinize the separation power of the variables, --- Cuts : and to remove the weakest ones. If some among the input variables --- Cuts : can be described by a single cut (e.g., because signal tends to be --- Cuts : larger than background), this can be indicated to MethodCuts via --- Cuts : the "Fsmart" options (see option string). Choosing this option --- Cuts : reduces the number of requirements for the variable from 2 (min/max) --- Cuts : to a single one (TMVA finds out whether it is to be interpreted as --- Cuts : min or max). --- Cuts : --- Cuts : [1m--- Performance tuning via configuration options:[0m --- Cuts : --- Cuts : Monte Carlo sampling: --- Cuts : --- Cuts : Apart form the "Fsmart" option for the variables, the only way --- Cuts : to improve the MC sampling is to increase the sampling rate. This --- Cuts : is done via the configuration option "MC_NRandCuts". The execution --- Cuts : time scales linearly with the sampling rate. --- Cuts : --- Cuts : Genetic Algorithm: --- Cuts : --- Cuts : The algorithm terminates if no significant fitness increase has --- Cuts : been achieved within the last "nsteps" steps of the calculation. --- Cuts : Wiggles in the ROC curve or constant background rejection of 1 --- Cuts : indicate that the GA failed to always converge at the true maximum --- Cuts : fitness. In such a case, it is recommended to broaden the search --- Cuts : by increasing the population size ("popSize") and to give the GA --- Cuts : more time to find improvements by increasing the number of steps --- Cuts : ("nsteps") --- Cuts : -> increase "popSize" (at least >10 * number of variables) --- Cuts : -> increase "nsteps" --- Cuts : --- Cuts : Simulated Annealing (SA) algorithm: --- Cuts : --- Cuts : "Increasing Adaptive" approach: --- Cuts : --- Cuts : The algorithm seeks local minima and explores their neighborhood, while --- Cuts : changing the ambient temperature depending on the number of failures --- Cuts : in the previous steps. The performance can be improved by increasing --- Cuts : the number of iteration steps ("MaxCalls"), or by adjusting the --- Cuts : minimal temperature ("MinTemperature"). Manual adjustments of the --- Cuts : speed of the temperature increase ("TemperatureScale" and "AdaptiveSpeed") --- Cuts : to individual data sets should also help. Summary: --- Cuts : -> increase "MaxCalls" --- Cuts : -> adjust "MinTemperature" --- Cuts : -> adjust "TemperatureScale" --- Cuts : -> adjust "AdaptiveSpeed" --- Cuts : --- Cuts : "Decreasing Adaptive" approach: --- Cuts : --- Cuts : The algorithm calculates the initial temperature (based on the effect- --- Cuts : iveness of large steps) and the multiplier that ensures to reach the --- Cuts : minimal temperature with the requested number of iteration steps. --- Cuts : The performance can be improved by adjusting the minimal temperature --- Cuts : ("MinTemperature") and by increasing number of steps ("MaxCalls"): --- Cuts : -> increase "MaxCalls" --- Cuts : -> adjust "MinTemperature" --- Cuts : --- Cuts : Other kernels: --- Cuts : --- Cuts : Alternative ways of counting the temperature change are implemented. --- Cuts : Each of them starts with the maximum temperature ("MaxTemperature") --- Cuts : and descreases while changing the temperature according to a given --- Cuts : prescription: --- Cuts : CurrentTemperature = --- Cuts : - Sqrt: InitialTemperature / Sqrt(StepNumber+2) * TemperatureScale --- Cuts : - Log: InitialTemperature / Log(StepNumber+2) * TemperatureScale --- Cuts : - Homo: InitialTemperature / (StepNumber+2) * TemperatureScale --- Cuts : - Sin: ( Sin( StepNumber / TemperatureScale ) + 1 ) / (StepNumber + 1) * InitialTemperature + Eps --- Cuts : - Geo: CurrentTemperature * TemperatureScale --- Cuts : --- Cuts : Their performance can be improved by adjusting initial temperature --- Cuts : ("InitialTemperature"), the number of iteration steps ("MaxCalls"), --- Cuts : and the multiplier that scales the termperature descrease --- Cuts : ("TemperatureScale") --- Cuts : -> increase "MaxCalls" --- Cuts : -> adjust "InitialTemperature" --- Cuts : -> adjust "TemperatureScale" --- Cuts : -> adjust "KernelTemperature" --- Cuts : --- Cuts : <Suppress this message by specifying "!H" in the booking option> --- Cuts : [1m================================================================[0m --- Cuts : --- CutsFitter_GA : Parsing option string: --- CutsFitter_GA : "!H:!V:!FitMethod=GA:!EffSel:Steps=30:Cycles=3:PopSize=100:SC_steps=10:SC_rate=5:SC_factor=0.95:!VarProp=FSmart" --- CutsFitter_GA : The following options are set: --- CutsFitter_GA : - By User: --- CutsFitter_GA : PopSize: "100" [Population size for GA] --- CutsFitter_GA : Steps: "30" [Number of steps for convergence] --- CutsFitter_GA : Cycles: "3" [Independent cycles of GA fitting] --- CutsFitter_GA : SC_steps: "10" [Spread control, steps] --- CutsFitter_GA : SC_rate: "5" [Spread control, rate: factor is changed depending on the rate] --- CutsFitter_GA : SC_factor: "0.95" [Spread control, factor] --- CutsFitter_GA : - Default: --- CutsFitter_GA : ConvCrit: "0.001" [Convergence criteria] --- CutsFitter_GA : SaveBestGen: "1" [Saves the best n results from each generation; these are included in the last cycle] --- CutsFitter_GA : SaveBestCycle: "10" [Saves the best n results from each cycle; these are included in the last cycle] --- CutsFitter_GA : Trim: "False" [Trim the population to PopSize after assessing the fitness of each individual] --- CutsFitter_GA : Seed: "100" [Set seed of random generator (0 gives random seeds)] --- CutsFitter_GA : <GeneticFitter> Optimisation, please be patient ... (note: inaccurate progress timing for GA) --- CutsFitter_GA : Elapsed time: [1;31m2.99 sec[0m --- Cuts : ----------------------------------------------------- --- Cuts : Cut values for requested signal efficiency: 0.1 --- Cuts : Corresponding background efficiency : 0.0108147) --- Cuts : ----------------------------------------------------- --- Cuts : Cut[ 0]: -1e+30 < [HT] <= 1642.11 --- Cuts : Cut[ 1]: -1e+30 < [Jet1Pt] <= 58.4718 --- Cuts : Cut[ 2]: 2.2199 < [DeltaRJet1Jet2] <= 1e+30 --- Cuts : Cut[ 3]: -1e+30 < [WTransverseMass] <= 282.28 --- Cuts : ----------------------------------------------------- --- Cuts : ----------------------------------------------------- --- Cuts : Cut values for requested signal efficiency: 0.2 --- Cuts : Corresponding background efficiency : 0.0197783) --- Cuts : ----------------------------------------------------- --- Cuts : Cut[ 0]: -1e+30 < [HT] <= 243.069 --- Cuts : Cut[ 1]: -1e+30 < [Jet1Pt] <= 565.519 --- Cuts : Cut[ 2]: 0.892931 < [DeltaRJet1Jet2] <= 1e+30 --- Cuts : Cut[ 3]: -1e+30 < [WTransverseMass] <= 288.73 --- Cuts : ----------------------------------------------------- --- Cuts : ----------------------------------------------------- --- Cuts : Cut values for requested signal efficiency: 0.3 --- Cuts : Corresponding background efficiency : 0.0357081) --- Cuts : ----------------------------------------------------- --- Cuts : Cut[ 0]: -1e+30 < [HT] <= 258.284 --- Cuts : Cut[ 1]: -1e+30 < [Jet1Pt] <= 331.66 --- Cuts : Cut[ 2]: 1.03693 < [DeltaRJet1Jet2] <= 1e+30 --- Cuts : Cut[ 3]: -1e+30 < [WTransverseMass] <= 343.833 --- Cuts : ----------------------------------------------------- --- Cuts : ----------------------------------------------------- --- Cuts : Cut values for requested signal efficiency: 0.4 --- Cuts : Corresponding background efficiency : 0.0574593) --- Cuts : ----------------------------------------------------- --- Cuts : Cut[ 0]: -1e+30 < [HT] <= 280.778 --- Cuts : Cut[ 1]: -1e+30 < [Jet1Pt] <= 459.721 --- Cuts : Cut[ 2]: 1.03693 < [DeltaRJet1Jet2] <= 1e+30 --- Cuts : Cut[ 3]: -1e+30 < [WTransverseMass] <= 98.7328 --- Cuts : ----------------------------------------------------- --- Cuts : ----------------------------------------------------- --- Cuts : Cut values for requested signal efficiency: 0.5 --- Cuts : Corresponding background efficiency : 0.100694) --- Cuts : ----------------------------------------------------- --- Cuts : Cut[ 0]: -1e+30 < [HT] <= 295.411 --- Cuts : Cut[ 1]: -1e+30 < [Jet1Pt] <= 1068.21 --- Cuts : Cut[ 2]: 0.787902 < [DeltaRJet1Jet2] <= 1e+30 --- Cuts : Cut[ 3]: -1e+30 < [WTransverseMass] <= 117.266 --- Cuts : ----------------------------------------------------- --- Cuts : ----------------------------------------------------- --- Cuts : Cut values for requested signal efficiency: 0.6 --- Cuts : Corresponding background efficiency : 0.177131) --- Cuts : ----------------------------------------------------- --- Cuts : Cut[ 0]: -1e+30 < [HT] <= 321.171 --- Cuts : Cut[ 1]: -1e+30 < [Jet1Pt] <= 460.868 --- Cuts : Cut[ 2]: 1.03693 < [DeltaRJet1Jet2] <= 1e+30 --- Cuts : Cut[ 3]: -1e+30 < [WTransverseMass] <= 157.666 --- Cuts : ----------------------------------------------------- --- Cuts : ----------------------------------------------------- --- Cuts : Cut values for requested signal efficiency: 0.7 --- Cuts : Corresponding background efficiency : 0.304926) --- Cuts : ----------------------------------------------------- --- Cuts : Cut[ 0]: -1e+30 < [HT] <= 356.563 --- Cuts : Cut[ 1]: -1e+30 < [Jet1Pt] <= 1036.81 --- Cuts : Cut[ 2]: 0.620059 < [DeltaRJet1Jet2] <= 1e+30 --- Cuts : Cut[ 3]: -1e+30 < [WTransverseMass] <= 280.334 --- Cuts : ----------------------------------------------------- --- Cuts : ----------------------------------------------------- --- Cuts : Cut values for requested signal efficiency: 0.8 --- Cuts : Corresponding background efficiency : 0.435049) --- Cuts : ----------------------------------------------------- --- Cuts : Cut[ 0]: -1e+30 < [HT] <= 2316.92 --- Cuts : Cut[ 1]: -1e+30 < [Jet1Pt] <= 122.678 --- Cuts : Cut[ 2]: 1.02789 < [DeltaRJet1Jet2] <= 1e+30 --- Cuts : Cut[ 3]: -1e+30 < [WTransverseMass] <= 115.868 --- Cuts : ----------------------------------------------------- --- Cuts : ----------------------------------------------------- --- Cuts : Cut values for requested signal efficiency: 0.9 --- Cuts : Corresponding background efficiency : 0.671083) --- Cuts : ----------------------------------------------------- --- Cuts : Cut[ 0]: -1e+30 < [HT] <= 515.313 --- Cuts : Cut[ 1]: -1e+30 < [Jet1Pt] <= 620.115 --- Cuts : Cut[ 2]: 0.887084 < [DeltaRJet1Jet2] <= 1e+30 --- Cuts : Cut[ 3]: -1e+30 < [WTransverseMass] <= 120.256 --- Cuts : ----------------------------------------------------- --- Cuts : Creating weight file in text format: [1;34mweights/TMVAnalysis_CutsGA.weights.txt[0m --- Cuts : Creating standalone response class : [1;34mweights/TMVAnalysis_CutsGA.class.C[0m --- Cuts : write monitoring histograms to file: TMVAout.root:/Method_Cuts/CutsGA --- Factory : Train method: Likelihood --- Likelihood : Filling reference histograms --- PDF : Validation result for PDF "HT signal training": --- PDF : chi2/ndof(!=0) = 39.4/7 = 5.63 (Prob = 0.00) --- PDF : #bins-found(#expected-bins) deviating > [1,2,3,6] sigmas: [3(2),3(0),2(0),0(0)] --- PDF : Validation result for PDF "Jet1Pt signal training": --- PDF : chi2/ndof(!=0) = 394.1/5 = 78.81 (Prob = 0.00) --- PDF : #bins-found(#expected-bins) deviating > [1,2,3,6] sigmas: [4(1),4(0),3(0),3(0)] --- PDF : Validation result for PDF "DeltaRJet1Jet2 signal training": --- PDF : chi2/ndof(!=0) = 121.7/9 = 13.52 (Prob = 0.00) --- PDF : #bins-found(#expected-bins) deviating > [1,2,3,6] sigmas: [7(2),4(0),3(0),2(0)] --- PDF : Validation result for PDF "WTransverseMass signal training": --- PDF : chi2/ndof(!=0) = 650.4/5 = 130.09 (Prob = 0.00) --- PDF : #bins-found(#expected-bins) deviating > [1,2,3,6] sigmas: [4(1),3(0),3(0),1(0)] --- PDF : Validation result for PDF "HT background training": --- PDF : chi2/ndof(!=0) = 200.5/8 = 25.06 (Prob = 0.00) --- PDF : #bins-found(#expected-bins) deviating > [1,2,3,6] sigmas: [6(2),4(0),4(0),3(0)] --- PDF : Validation result for PDF "Jet1Pt background training": --- PDF : chi2/ndof(!=0) = 33.0/7 = 4.71 (Prob = 0.00) --- PDF : #bins-found(#expected-bins) deviating > [1,2,3,6] sigmas: [4(2),3(0),2(0),0(0)] --- PDF : Validation result for PDF "DeltaRJet1Jet2 background training": --- PDF : chi2/ndof(!=0) = 164.4/9 = 18.27 (Prob = 0.00) --- PDF : #bins-found(#expected-bins) deviating > [1,2,3,6] sigmas: [6(2),6(0),5(0),2(0)] --- PDF : Validation result for PDF "WTransverseMass background training": --- PDF : chi2/ndof(!=0) = 295.0/8 = 36.87 (Prob = 0.00) --- PDF : #bins-found(#expected-bins) deviating > [1,2,3,6] sigmas: [6(2),4(0),3(0),2(0)] --- Likelihood : Creating weight file in text format: [1;34mweights/TMVAnalysis_Likelihood.weights.txt[0m --- Likelihood : Creating standalone response class : [1;34mweights/TMVAnalysis_Likelihood.class.C[0m --- Likelihood : Write monitoring histograms to file: TMVAout.root:/Method_Likelihood/Likelihood --- Factory : Train method: LikelihoodPCA --- Likelihood : Filling reference histograms --- PDF : Validation result for PDF "HT signal training": --- PDF : chi2/ndof(!=0) = 60.4/7 = 8.62 (Prob = 0.00) --- PDF : #bins-found(#expected-bins) deviating > [1,2,3,6] sigmas: [4(2),3(0),3(0),0(0)] --- PDF : Validation result for PDF "Jet1Pt signal training": --- PDF : chi2/ndof(!=0) = 454.2/8 = 56.77 (Prob = 0.00) --- PDF : #bins-found(#expected-bins) deviating > [1,2,3,6] sigmas: [5(2),4(0),3(0),3(0)] --- PDF : Validation result for PDF "DeltaRJet1Jet2 signal training": --- PDF : chi2/ndof(!=0) = 800.4/5 = 160.09 (Prob = 0.00) --- PDF : #bins-found(#expected-bins) deviating > [1,2,3,6] sigmas: [5(1),5(0),4(0),4(0)] --- PDF : Validation result for PDF "WTransverseMass signal training": --- PDF : chi2/ndof(!=0) = 95.2/10 = 9.52 (Prob = 0.00) --- PDF : #bins-found(#expected-bins) deviating > [1,2,3,6] sigmas: [7(3),7(0),5(0),0(0)] --- PDF : Validation result for PDF "HT background training": --- PDF : chi2/ndof(!=0) = 10274.0/3 = 3424.65 (Prob = 0.00) --- PDF : #bins-found(#expected-bins) deviating > [1,2,3,6] sigmas: [2(0),2(0),2(0),2(0)] --- PDF : Validation result for PDF "Jet1Pt background training": --- PDF : chi2/ndof(!=0) = 227.4/8 = 28.43 (Prob = 0.00) --- PDF : #bins-found(#expected-bins) deviating > [1,2,3,6] sigmas: [5(2),4(0),4(0),3(0)] --- PDF : Validation result for PDF "DeltaRJet1Jet2 background training": --- PDF : chi2/ndof(!=0) = 297.4/8 = 37.17 (Prob = 0.00) --- PDF : #bins-found(#expected-bins) deviating > [1,2,3,6] sigmas: [5(2),4(0),4(0),4(0)] --- PDF : Validation result for PDF "WTransverseMass background training": --- PDF : chi2/ndof(!=0) = 97.3/8 = 12.16 (Prob = 0.00) --- PDF : #bins-found(#expected-bins) deviating > [1,2,3,6] sigmas: [7(2),5(0),5(0),0(0)] --- Likelihood : Creating weight file in text format: [1;34mweights/TMVAnalysis_LikelihoodPCA.weights.txt[0m --- Likelihood : Creating standalone response class : [1;34mweights/TMVAnalysis_LikelihoodPCA.class.C[0m --- Likelihood : Write monitoring histograms to file: TMVAout.root:/Method_Likelihood/LikelihoodPCA --- Factory : Train method: PDERS --- PDERS : Creating weight file in text format: [1;34mweights/TMVAnalysis_PDERS.weights.txt[0m --- PDERS : Creating standalone response class : [1;34mweights/TMVAnalysis_PDERS.class.C[0m --- PDERS : No monitoring histograms written --- Factory : Train method: KNN --- KNN : <Train> start... --- KNN : Reading 1339 events --- KNN : Number of signal events 515.581 --- KNN : Number of background events 824.079 --- KNN : Creating kd-tree with 1339 events --- ModulekNN : Computing scale factor for 1d distributions: (ifrac, bottom, top) = (80%, 10%, 90%) --- ModulekNN : Optimizing tree for 4 variables with 1339 values --- ModulekNN : <Fill> Class 1 has 516 events --- ModulekNN : <Fill> Class 2 has 823 events --- KNN : Creating weight file in text format: [1;34mweights/TMVAnalysis_KNN.weights.txt[0m --- KNN : Starting WriteWeightsToStream(ostream& os) function... --- KNN : Creating standalone response class : [1;34mweights/TMVAnalysis_KNN.class.C[0m --- KNN : No monitoring histograms written --- Factory : Train method: HMatrix --- HMatrix : Creating weight file in text format: [1;34mweights/TMVAnalysis_HMatrix.weights.txt[0m --- HMatrix : Creating standalone response class : [1;34mweights/TMVAnalysis_HMatrix.class.C[0m --- HMatrix : No monitoring histograms written --- Factory : Train method: Fisher --- Fisher : --- Fisher : [1m================================================================[0m --- Fisher : [1mH e l p f o r c l a s s i f i e r [ Fisher ] :[0m --- Fisher : --- Fisher : [1m--- Short description:[0m --- Fisher : --- Fisher : Fisher discriminants select events by distinguishing the mean --- Fisher : values of the signal and background distributions in a trans- --- Fisher : formed variable space where linear correlations are removed. --- Fisher : --- Fisher : (More precisely: the "linear discriminator" determines --- Fisher : an axis in the (correlated) hyperspace of the input --- Fisher : variables such that, when projecting the output classes --- Fisher : (signal and background) upon this axis, they are pushed --- Fisher : as far as possible away from each other, while events --- Fisher : of a same class are confined in a close vicinity. The --- Fisher : linearity property of this classifier is reflected in the --- Fisher : metric with which "far apart" and "close vicinity" are --- Fisher : determined: the covariance matrix of the discriminating --- Fisher : variable space.) --- Fisher : --- Fisher : [1m--- Performance optimisation:[0m --- Fisher : --- Fisher : Optimal performance for Fisher discriminants is obtained for --- Fisher : linearly correlated Gaussian-distributed variables. Any deviation --- Fisher : from this ideal reduces the achievable separation power. In --- Fisher : particular, no discrimination at all is achieved for a variable --- Fisher : that has the same sample mean for signal and background, even if --- Fisher : the shapes of the distributions are very different. Thus, Fisher --- Fisher : discriminants often benefit from suitable transformations of the --- Fisher : input variables. For example, if a variable x in [-1,1] has a --- Fisher : a parabolic signal distributions, and a uniform background --- Fisher : distributions, their mean value is zero in both cases, leading --- Fisher : to no separation. The simple transformation x -> |x| renders this --- Fisher : variable powerful for the use in a Fisher discriminant. --- Fisher : --- Fisher : [1m--- Performance tuning via configuration options:[0m --- Fisher : --- Fisher : None --- Fisher : --- Fisher : <Suppress this message by specifying "!H" in the booking option> --- Fisher : [1m================================================================[0m --- Fisher : --- Fisher : Results for Fisher coefficients: --- Fisher : --------------------------------- --- Fisher : Variable: Coefficient: --- Fisher : --------------------------------- --- Fisher : HT: -0.002 --- Fisher : Jet1Pt: +0.001 --- Fisher : DeltaRJet1Jet2: +0.104 --- Fisher : WTransverseMass: +0.001 --- Fisher : (offset): +0.303 --- Fisher : --------------------------------- --- Fisher : <CreateMVAPdfs> Using 50 bins and smooth 1 times --- PDF : Validation result for PDF "Fisher_tr_S": --- PDF : chi2/ndof(!=0) = 3.5/25 = 0.14 (Prob = 1.00) --- PDF : #bins-found(#expected-bins) deviating > [1,2,3,6] sigmas: [0(7),0(1),0(0),0(0)] --- PDF : Validation result for PDF "Fisher_tr_B": --- PDF : chi2/ndof(!=0) = 7.0/32 = 0.22 (Prob = 1.00) --- PDF : #bins-found(#expected-bins) deviating > [1,2,3,6] sigmas: [2(10),0(1),0(0),0(0)] --- Fisher : <CreateMVAPdfs> Separation from histogram (PDF): 0.266 (0.256) --- Fisher : Creating weight file in text format: [1;34mweights/TMVAnalysis_Fisher.weights.txt[0m --- Fisher : Creating standalone response class : [1;34mweights/TMVAnalysis_Fisher.class.C[0m --- Fisher : No monitoring histograms written --- Factory : Train method: FDA_MT --- FDA : --- FDA : [1m================================================================[0m --- FDA : [1mH e l p f o r c l a s s i f i e r [ FDA ] :[0m --- FDA : --- FDA : [1m--- Short description:[0m --- FDA : --- FDA : The function discriminant analysis (FDA) is a classifier suitable --- FDA : to solve linear or simple nonlinear discrimination problems. --- FDA : --- FDA : The user provides the desired function with adjustable parameters --- FDA : via the configuration option string, and FDA fits the parameters to --- FDA : it, requiring the signal (background) function value to be as close --- FDA : as possible to 1 (0). Its advantage over the more involved and --- FDA : automatic nonlinear discriminators is the simplicity and transparency --- FDA : of the discrimination expression. A shortcoming is that FDA will --- FDA : underperform for involved problems with complicated, phase space --- FDA : dependent nonlinear correlations. --- FDA : --- FDA : Please consult the Users Guide for the format of the formula string --- FDA : and the allowed parameter ranges: --- FDA : http://tmva.sourceforge.net/docu/TMVAUsersGuide.pdf --- FDA : --- FDA : [1m--- Performance optimisation:[0m --- FDA : --- FDA : The FDA performance depends on the complexity and fidelity of the --- FDA : user-defined discriminator function. As a general rule, it should --- FDA : be able to reproduce the discrimination power of any linear --- FDA : discriminant analysis. To reach into the nonlinear domain, it is --- FDA : useful to inspect the correlation profiles of the input variables, --- FDA : and add quadratic and higher polynomial terms between variables as --- FDA : necessary. Comparison with more involved nonlinear classifiers can --- FDA : be used as a guide. --- FDA : --- FDA : [1m--- Performance tuning via configuration options:[0m --- FDA : --- FDA : Depending on the function used, the choice of "FitMethod" is --- FDA : crucial for getting valuable solutions with FDA. As a guideline it --- FDA : is recommended to start with "FitMethod=MINUIT". When more complex --- FDA : functions are used where MINUIT does not converge to reasonable --- FDA : results, the user should switch to non-gradient FitMethods such --- FDA : as GeneticAlgorithm (GA) or Monte Carlo (MC). It might prove to be --- FDA : useful to combine GA (or MC) with MINUIT by setting the option --- FDA : "Converger=MINUIT". GA (MC) will then set the starting parameters --- FDA : for MINUIT such that the basic quality of GA (MC) of finding global --- FDA : minima is combined with the efficacy of MINUIT of finding local --- FDA : minima. --- FDA : --- FDA : <Suppress this message by specifying "!H" in the booking option> --- FDA : [1m================================================================[0m --- FDA : --- FDA : Results for parameter fit using "MINUIT" fitter: --- FDA : ----------------------- --- FDA : Parameter: Fit result: --- FDA : ----------------------- --- FDA : Par(0): 0.694601 --- FDA : Par(1): -0.0011554 --- FDA : Par(2): 0.000436835 --- FDA : Par(3): 0.0637497 --- FDA : Par(4): 0.000475737 --- FDA : ----------------------- --- FDA : Discriminator expression: "(0)+(1)*x0+(2)*x1+(3)*x2+(4)*x3" --- FDA : Value of estimator at minimum: 0.43485 --- FDA : Creating weight file in text format: [1;34mweights/TMVAnalysis_FDA_MT.weights.txt[0m --- FDA : Creating standalone response class : [1;34mweights/TMVAnalysis_FDA_MT.class.C[0m --- FDA : No monitoring histograms written --- Factory : Train method: MLP --- MLP : --- MLP : [1m================================================================[0m --- MLP : [1mH e l p f o r c l a s s i f i e r [ MLP ] :[0m --- MLP : --- MLP : [1m--- Short description:[0m --- MLP : --- MLP : The MLP artificial neural network (ANN) is a traditional feed- --- MLP : forward multilayer perceptron impementation. The MLP has a user- --- MLP : defined hidden layer architecture, while the number of input (output) --- MLP : nodes is determined by the input variables (output classes, i.e., --- MLP : signal and one background). --- MLP : --- MLP : [1m--- Performance optimisation:[0m --- MLP : --- MLP : Neural networks are stable and performing for a large variety of --- MLP : linear and non-linear classification problems. However, in contrast --- MLP : to (e.g.) boosted decision trees, the user is advised to reduce the --- MLP : number of input variables that have only little discrimination power. --- MLP : --- MLP : In the tests we have carried out so far, the MLP and ROOT networks --- MLP : (TMlpANN, interfaced via TMVA) performed equally well, with however --- MLP : a clear speed advantage for the MLP. The Clermont-Ferrand neural --- MLP : net (CFMlpANN) exhibited worse classification performance in these --- MLP : tests, which is partly due to the slow convergence of its training --- MLP : (at least 10k training cycles are required to achieve approximately --- MLP : competitive results). --- MLP : --- MLP : [1mOvertraining: [0monly the TMlpANN performs an explicit separation of the --- MLP : full training sample into independent training and validation samples. --- MLP : We have found that in most high-energy physics applications the --- MLP : avaliable degrees of freedom (training events) are sufficient to --- MLP : constrain the weights of the relatively simple architectures required --- MLP : to achieve good performance. Hence no overtraining should occur, and --- MLP : the use of validation samples would only reduce the available training --- MLP : information. However, if the perrormance on the training sample is --- MLP : found to be significantly better than the one found with the inde- --- MLP : pendent test sample, caution is needed. The results for these samples --- MLP : are printed to standard output at the end of each training job. --- MLP : --- MLP : [1m--- Performance tuning via configuration options:[0m --- MLP : --- MLP : The hidden layer architecture for all ANNs is defined by the option --- MLP : "HiddenLayers=N+1,N,...", where here the first hidden layer has N+1 --- MLP : neurons and the second N neurons (and so on), and where N is the number --- MLP : of input variables. Excessive numbers of hidden layers should be avoided, --- MLP : in favour of more neurons in the first hidden layer. --- MLP : --- MLP : The number of cycles should be above 500. As said, if the number of --- MLP : adjustable weights is small compared to the training sample size, --- MLP : using a large number of training samples should not lead to overtraining. --- MLP : --- MLP : <Suppress this message by specifying "!H" in the booking option> --- MLP : [1m================================================================[0m --- MLP : --- MLP : Training Network --- MLP : Train: elapsed time: [1;31m5.19 sec[0m --- MLP : Creating weight file in text format: [1;34mweights/TMVAnalysis_MLP.weights.txt[0m --- MLP : Creating standalone response class : [1;34mweights/TMVAnalysis_MLP.class.C[0m --- MLP : Write special histos to file: TMVAout.root:/Method_MLP/MLP --- Factory : Train method: SVM_Gauss --- SVM : Sorry, no computing time forecast available for SVM, please wait ... --- SVM : <Train> elapsed time: [1;31m0.0389 sec[0m --- SVM : <Train> number of iterations: 5 --- SVM : Results: --- SVM : - number of support vectors: 9 (0%) --- SVM : - b: 0.356742 --- SVM : All support vectors stored properly --- SVM : Creating weight file in text format: [1;34mweights/TMVAnalysis_SVM_Gauss.weights.txt[0m --- SVM : Creating standalone response class : [1;34mweights/TMVAnalysis_SVM_Gauss.class.C[0m --- SVM : No monitoring histograms written --- Factory : Train method: BDT --- BDT : Training 400 Decision Trees ... patience please --- BDT : <Train> elapsed time: [1;31m29.9 sec[0m --- BDT : <Train> average number of nodes before/after pruning : 100 / 32 --- BDT : Creating weight file in text format: [1;34mweights/TMVAnalysis_BDT.weights.txt[0m --- BDT : Creating standalone response class : [1;34mweights/TMVAnalysis_BDT.class.C[0m --- BDT : Write monitoring histograms to file: TMVAout.root:/Method_BDT/BDT --- Factory : Train method: BDTD --- BDT : Training 400 Decision Trees ... patience please --- BDT : <Train> elapsed time: [1;31m30.5 sec[0m --- BDT : <Train> average number of nodes before/after pruning : 100 / 37 --- BDT : Creating weight file in text format: [1;34mweights/TMVAnalysis_BDTD.weights.txt[0m --- BDT : Creating standalone response class : [1;34mweights/TMVAnalysis_BDTD.class.C[0m --- BDT : Write monitoring histograms to file: TMVAout.root:/Method_BDT/BDTD --- Factory : Train method: RuleFit --- RuleFit : --- RuleFit : [1m================================================================[0m --- RuleFit : [1mH e l p f o r c l a s s i f i e r [ RuleFit ] :[0m --- RuleFit : --- RuleFit : [1m--- Short description:[0m --- RuleFit : --- RuleFit : This method uses a collection of so called rules to create a --- RuleFit : discriminating scoring function. Each rule consists of a series --- RuleFit : of cuts in parameter space. The ensemble of rules are created --- RuleFit : from a forest of decision trees, trained using the training data. --- RuleFit : Each node (apart from the root) corresponds to one rule. --- RuleFit : The scoring function is then obtained by linearly combining --- RuleFit : the rules. A fitting procedure is applied to find the optimum --- RuleFit : set of coefficients. The goal is to find a model with few rules --- RuleFit : but with a strong discriminating power. --- RuleFit : --- RuleFit : [1m--- Performance optimisation:[0m --- RuleFit : --- RuleFit : There are two important considerations to make when optimising: --- RuleFit : --- RuleFit : 1. Topology of the decision tree forest --- RuleFit : 2. Fitting of the coefficients --- RuleFit : --- RuleFit : The maximum complexity of the rules is defined by the size of --- RuleFit : the trees. Large trees will yield many complex rules and capture --- RuleFit : higher order correlations. On the other hand, small trees will --- RuleFit : lead to a smaller ensemble with simple rules, only capable of --- RuleFit : modeling simple structures. --- RuleFit : Several parameters exists for controlling the complexity of the --- RuleFit : rule ensemble. --- RuleFit : --- RuleFit : The fitting procedure searches for a minimum using a gradient --- RuleFit : directed path. Apart from step size and number of steps, the --- RuleFit : evolution of the path is defined by a cut-off parameter, tau. --- RuleFit : This parameter is unknown and depends on the training data. --- RuleFit : A large value will tend to give large weights to a few rules. --- RuleFit : Similarily, a small value will lead to a large set of rules --- RuleFit : with similar weights. --- RuleFit : --- RuleFit : A final point is the model used; rules and/or linear terms. --- RuleFit : For a given training sample, the result may improve by adding --- RuleFit : linear terms. If best performance is optained using only linear --- RuleFit : terms, it is very likely that the Fisher discriminant would be --- RuleFit : a better choice. Ideally the fitting procedure should be able to --- RuleFit : make this choice by giving appropriate weights for either terms. --- RuleFit : --- RuleFit : [1m--- Performance tuning via configuration options:[0m --- RuleFit : --- RuleFit : I. TUNING OF RULE ENSEMBLE: --- RuleFit : --- RuleFit : [1mForestType [0m: Recomended is to use the default "AdaBoost". --- RuleFit : [1mnTrees [0m: More trees leads to more rules but also slow --- RuleFit : performance. With too few trees the risk is --- RuleFit : that the rule ensemble becomes too simple. --- RuleFit : [1mfEventsMin [0m --- RuleFit : [1mfEventsMax [0m: With a lower min, more large trees will be generated --- RuleFit : leading to more complex rules. --- RuleFit : With a higher max, more small trees will be --- RuleFit : generated leading to more simple rules. --- RuleFit : By changing this range, the average complexity --- RuleFit : of the rule ensemble can be controlled. --- RuleFit : [1mRuleMinDist [0m: By increasing the minimum distance between --- RuleFit : rules, fewer and more diverse rules will remain. --- RuleFit : Initially it is a good idea to keep this small --- RuleFit : or zero and let the fitting do the selection of --- RuleFit : rules. In order to reduce the ensemble size, --- RuleFit : the value can then be increased. --- RuleFit : --- RuleFit : II. TUNING OF THE FITTING: --- RuleFit : --- RuleFit : [1mGDPathEveFrac [0m: fraction of events in path evaluation --- RuleFit : Increasing this fraction will improve the path --- RuleFit : finding. However, a too high value will give few --- RuleFit : unique events available for error estimation. --- RuleFit : It is recomended to usethe default = 0.5. --- RuleFit : [1mGDTau [0m: cutoff parameter tau --- RuleFit : By default this value is set to -1.0. --- RuleFit : This means that the cut off parameter is --- RuleFit : automatically estimated. In most cases --- RuleFit : this should be fine. However, you may want --- RuleFit : to fix this value if you already know it --- RuleFit : and want to reduce on training time. --- RuleFit : [1mGDTauPrec [0m: precision of estimated tau --- RuleFit : Increase this precision to find a more --- RuleFit : optimum cut-off parameter. --- RuleFit : [1mGDNStep [0m: number of steps in path search --- RuleFit : If the number of steps is too small, then --- RuleFit : the program will give a warning message. --- RuleFit : --- RuleFit : III. WARNING MESSAGES --- RuleFit : --- RuleFit : [1mRisk(i+1)>=Risk(i) in path[0m --- RuleFit : [1mChaotic behaviour of risk evolution.[0m --- RuleFit : The error rate was still decreasing at the end --- RuleFit : By construction the Risk should always decrease. --- RuleFit : However, if the training sample is too small or --- RuleFit : the model is overtrained, such warnings can --- RuleFit : occur. --- RuleFit : The warnings can safely be ignored if only a --- RuleFit : few (<3) occur. If more warnings are generated, --- RuleFit : the fitting fails. --- RuleFit : A remedy may be to increase the value --- RuleFit : [1mGDValidEveFrac[0m to 1.0 (or a larger value). --- RuleFit : In addition, if [1mGDPathEveFrac[0m is too high --- RuleFit : the same warnings may occur since the events --- RuleFit : used for error estimation are also used for --- RuleFit : path estimation. --- RuleFit : Another possibility is to modify the model - --- RuleFit : See above on tuning the rule ensemble. --- RuleFit : --- RuleFit : [1mThe error rate was still decreasing at the end of the path[0m --- RuleFit : Too few steps in path! Increase [1mGDNSteps[0m. --- RuleFit : --- RuleFit : [1mReached minimum early in the search[0m --- RuleFit : Minimum was found early in the fitting. This --- RuleFit : may indicate that the used step size [1mGDStep[0m. --- RuleFit : was too large. Reduce it and rerun. --- RuleFit : If the results still are not OK, modify the --- RuleFit : model either by modifying the rule ensemble --- RuleFit : or add/remove linear terms --- RuleFit : --- RuleFit : <Suppress this message by specifying "!H" in the booking option> --- RuleFit : [1m================================================================[0m --- RuleFit : --- RuleFit : -------------------RULE ENSEMBLE SUMMARY------------------------ --- RuleFit : Tree training method : AdaBoost --- RuleFit : Number of events per tree : 1339 --- RuleFit : Number of trees : 20 --- RuleFit : Number of generated rules : 242 --- RuleFit : Idem, after cleanup : 177 --- RuleFit : Average number of cuts per rule : 3.59 --- RuleFit : Spread in number of cuts per rules : 1.84 --- RuleFit : ---------------------------------------------------------------- --- RuleFit : --- RuleFit : GD path scan - the scan stops when the max num. of steps is reached or a min is found --- RuleFit : Estimating the cutoff parameter tau. The estimated time is a pessimistic maximum. --- RuleFit : Best path found with tau = 0.9293 after [1;31m3.16 sec[0m --- RuleFit : Fitting model... --- RuleFit : Minimization elapsed time : [1;31m1.82 sec[0m --- RuleFit : ---------------------------------------------------------------- --- RuleFit : Found minimum at step 10000 with error = 1.49845 --- RuleFit : Reason for ending loop: end of loop reached --- RuleFit : ---------------------------------------------------------------- [1;31m--- <WARNING> RuleFit : The error rate was still decreasing at the end of the path[0m [1;31m--- <WARNING> RuleFit : Increase number of steps (GDNSteps).[0m --- RuleFit : Elapsed time: [1;31m5.46 sec[0m --- RuleFit : Removed 175 out of a total of 177 rules with importance < 0.001 --- RuleFit : --- RuleFit : ================================================================ --- RuleFit : M o d e l --- RuleFit : ================================================================ --- RuleFit : Offset (a0) = 0.00921949 --- RuleFit : ------------------------------------ --- RuleFit : Linear model (weights unnormalised) --- RuleFit : ------------------------------------ --- RuleFit : Variable : Weights : Importance --- RuleFit : ------------------------------------ --- RuleFit : HT : -6.727e-03 : 1.000 --- RuleFit : Jet1Pt-> importance below threshhold = 0.000 --- RuleFit : DeltaRJet1Jet2 : -1.266e-02 : 0.008 --- RuleFit : WTransverseMass-> importance below threshhold = 0.000 --- RuleFit : ------------------------------------ --- RuleFit : Number of rules = 2 --- RuleFit : Printing the first 2 rules, ordered in importance. --- RuleFit : Rule 1 : Importance = 0.0028 --- RuleFit : Cut 1 : 304 < HT --- RuleFit : Rule 2 : Importance = 0.0028 --- RuleFit : Cut 1 : 310 < HT --- RuleFit : All rules printed --- RuleFit : ================================================================ --- RuleFit : --- RuleFit : Creating weight file in text format: [1;34mweights/TMVAnalysis_RuleFit.weights.txt[0m --- RuleFit : Creating standalone response class : [1;34mweights/TMVAnalysis_RuleFit.class.C[0m --- RuleFit : write monitoring ntuple to file: TMVAout.root:/Method_RuleFit/RuleFit --- Factory : --- Factory : Begin ranking of input variables... --- Factory : No variable ranking supplied by classifier: CutsGA --- Likelihood : Ranking result (top variable is best ranked) --- Likelihood : ---------------------------------------------------------------- --- Likelihood : Rank : Variable : Delta Separation --- Likelihood : ---------------------------------------------------------------- --- Likelihood : 1 : HT : 1.573e-01 --- Likelihood : 2 : Jet1Pt : 2.590e-03 --- Likelihood : 3 : DeltaRJet1Jet2 : -1.601e-02 --- Likelihood : 4 : WTransverseMass : -5.441e-02 --- Likelihood : ---------------------------------------------------------------- --- Likelihood : Ranking result (top variable is best ranked) --- Likelihood : ---------------------------------------------------------------- --- Likelihood : Rank : Variable : Delta Separation --- Likelihood : ---------------------------------------------------------------- --- Likelihood : 1 : HT : 2.026e-01 --- Likelihood : 2 : Jet1Pt : -1.315e-02 --- Likelihood : 3 : DeltaRJet1Jet2 : -2.915e-02 --- Likelihood : 4 : WTransverseMass : -3.787e-02 --- Likelihood : ---------------------------------------------------------------- --- Factory : No variable ranking supplied by classifier: PDERS --- Factory : No variable ranking supplied by classifier: KNN --- Factory : No variable ranking supplied by classifier: HMatrix --- Fisher : Ranking result (top variable is best ranked) --- Fisher : ---------------------------------------------------------------- --- Fisher : Rank : Variable : Discr. power --- Fisher : ---------------------------------------------------------------- --- Fisher : 1 : HT : 8.467e-02 --- Fisher : 2 : Jet1Pt : 7.024e-02 --- Fisher : 3 : DeltaRJet1Jet2 : 3.259e-03 --- Fisher : 4 : WTransverseMass : 6.449e-04 --- Fisher : ---------------------------------------------------------------- --- Factory : No variable ranking supplied by classifier: FDA_MT --- MLP : Ranking result (top variable is best ranked) --- MLP : ---------------------------------------------------------------- --- MLP : Rank : Variable : Importance --- MLP : ---------------------------------------------------------------- --- MLP : 1 : HT : 1.705e-01 --- MLP : 2 : DeltaRJet1Jet2 : 5.423e-02 --- MLP : 3 : Jet1Pt : 3.493e-02 --- MLP : 4 : WTransverseMass : 5.648e-05 --- MLP : ---------------------------------------------------------------- --- Factory : No variable ranking supplied by classifier: SVM_Gauss --- BDT : Ranking result (top variable is best ranked) --- BDT : ---------------------------------------------------------------- --- BDT : Rank : Variable : Variable Importance --- BDT : ---------------------------------------------------------------- --- BDT : 1 : DeltaRJet1Jet2 : 2.834e-01 --- BDT : 2 : HT : 2.664e-01 --- BDT : 3 : WTransverseMass : 2.550e-01 --- BDT : 4 : Jet1Pt : 1.952e-01 --- BDT : ---------------------------------------------------------------- --- BDT : Ranking result (top variable is best ranked) --- BDT : ---------------------------------------------------------------- --- BDT : Rank : Variable : Variable Importance --- BDT : ---------------------------------------------------------------- --- BDT : 1 : HT : 3.294e-01 --- BDT : 2 : DeltaRJet1Jet2 : 2.423e-01 --- BDT : 3 : WTransverseMass : 2.242e-01 --- BDT : 4 : Jet1Pt : 2.040e-01 --- BDT : ---------------------------------------------------------------- --- RuleFit : Ranking result (top variable is best ranked) --- RuleFit : ---------------------------------------------------------------- --- RuleFit : Rank : Variable : Importance --- RuleFit : ---------------------------------------------------------------- --- RuleFit : 1 : HT : 1.000e+00 --- RuleFit : 2 : DeltaRJet1Jet2 : 8.313e-03 --- RuleFit : 3 : Jet1Pt : 0.000e+00 --- RuleFit : 4 : WTransverseMass : 0.000e+00 --- RuleFit : ---------------------------------------------------------------- --- Factory : --- Factory : Testing all classifiers... --- Factory : Test method: CutsGA --- Cuts : Reading weight file: [1;34mweights/TMVAnalysis_CutsGA.weights.txt[0m --- Cuts : Read method with name <Cuts> and title <CutsGA> --- Cuts : Classifier was trained with TMVA Version: 3.9.5 --- Cuts : Classifier was trained with ROOT Version: 5.20/00 --- Cuts : Create VariableTransformation "None" --- Cuts : Use optimization method: 'Genetic Algorithm' --- Cuts : Use efficiency computation method: 'Event Selection' --- Cuts : Use "FSmart" cuts for variable: 'HT' --- Cuts : Use "FSmart" cuts for variable: 'Jet1Pt' --- Cuts : Use "FSmart" cuts for variable: 'DeltaRJet1Jet2' --- Cuts : Use "FSmart" cuts for variable: 'WTransverseMass' --- Cuts : Option for variable: HT: 'ForceSmart' (#: 3) --- Cuts : Option for variable: Jet1Pt: 'ForceSmart' (#: 3) --- Cuts : Option for variable: DeltaRJet1Jet2: 'ForceSmart' (#: 3) --- Cuts : Option for variable: WTransverseMass: 'ForceSmart' (#: 3) --- Cuts : Read cuts optimised using Genetic Algorithm --- Cuts : in 100 signal efficiency bins and for 4 variables --- Cuts : Preparing evaluation tree... --- Cuts : Elapsed time for evaluation of 4 events: [1;31m0.378 sec[0m --- Factory : Test method: Likelihood --- Likelihood : Reading weight file: [1;34mweights/TMVAnalysis_Likelihood.weights.txt[0m --- Likelihood : Read method with name <Likelihood> and title <Likelihood> --- Likelihood : Classifier was trained with TMVA Version: 3.9.5 --- Likelihood : Classifier was trained with ROOT Version: 5.20/00 --- Likelihood : Create VariableTransformation "None" --- Likelihood : Preparing evaluation tree... --- Likelihood : Elapsed time for evaluation of 4 events: [1;31m0.452 sec[0m --- Factory : Test method: LikelihoodPCA --- Likelihood : Reading weight file: [1;34mweights/TMVAnalysis_LikelihoodPCA.weights.txt[0m --- Likelihood : Read method with name <Likelihood> and title <LikelihoodPCA> --- Likelihood : Classifier was trained with TMVA Version: 3.9.5 --- Likelihood : Classifier was trained with ROOT Version: 5.20/00 --- Likelihood : Create VariableTransformation "PCA" --- Likelihood : Use principal component transformation --- Likelihood : Preparing evaluation tree... --- Likelihood : Elapsed time for evaluation of 4 events: [1;31m0.417 sec[0m --- Factory : Test method: PDERS --- PDERS : Reading weight file: [1;34mweights/TMVAnalysis_PDERS.weights.txt[0m --- PDERS : Read method with name <PDERS> and title <PDERS> --- PDERS : Classifier was trained with TMVA Version: 3.9.5 --- PDERS : Classifier was trained with ROOT Version: 5.20/00 --- PDERS : Create VariableTransformation "None" --- PDERS : Preparing evaluation tree... --- PDERS : Elapsed time for evaluation of 4 events: [1;31m0.356 sec[0m --- Factory : Test method: KNN --- KNN : Reading weight file: [1;34mweights/TMVAnalysis_KNN.weights.txt[0m --- KNN : Read method with name <KNN> and title <KNN> --- KNN : Classifier was trained with TMVA Version: 3.9.5 --- KNN : Classifier was trained with ROOT Version: 5.20/00 --- KNN : Create VariableTransformation "None" --- KNN : Starting ReadWeightsFromStream(istream& is) function... --- KNN : Erasing 1339 previously stored events --- KNN : Read 1339 events from text file --- KNN : Creating kd-tree with 1339 events --- ModulekNN : Computing scale factor for 1d distributions: (ifrac, bottom, top) = (80%, 10%, 90%) --- ModulekNN : Optimizing tree for 4 variables with 1339 values --- ModulekNN : <Fill> Class 1 has 516 events --- ModulekNN : <Fill> Class 2 has 823 events --- KNN : Preparing evaluation tree... --- KNN : Elapsed time for evaluation of 4 events: [1;31m0.499 sec[0m --- Factory : Test method: HMatrix --- HMatrix : Reading weight file: [1;34mweights/TMVAnalysis_HMatrix.weights.txt[0m --- HMatrix : Read method with name <HMatrix> and title <HMatrix> --- HMatrix : Classifier was trained with TMVA Version: 3.9.5 --- HMatrix : Classifier was trained with ROOT Version: 5.20/00 --- HMatrix : Create VariableTransformation "None" --- HMatrix : Preparing evaluation tree... --- HMatrix : Elapsed time for evaluation of 4 events: [1;31m0.387 sec[0m --- Factory : Test method: Fisher --- Fisher : Reading weight file: [1;34mweights/TMVAnalysis_Fisher.weights.txt[0m --- Fisher : Read method with name <Fisher> and title <Fisher> --- Fisher : Classifier was trained with TMVA Version: 3.9.5 --- Fisher : Classifier was trained with ROOT Version: 5.20/00 --- Fisher : Create VariableTransformation "None" --- Fisher : Preparing evaluation tree... --- Fisher : Elapsed time for evaluation of 4 events: [1;31m0.397 sec[0m --- Factory : Test method: FDA_MT --- FDA : Reading weight file: [1;34mweights/TMVAnalysis_FDA_MT.weights.txt[0m --- FDA : Read method with name <FDA> and title <FDA_MT> --- FDA : Classifier was trained with TMVA Version: 3.9.5 --- FDA : Classifier was trained with ROOT Version: 5.20/00 --- FDA : Create VariableTransformation "None" --- FDA : User-defined formula string : "(0)+(1)*x0+(2)*x1+(3)*x2+(4)*x3" --- FDA : TFormula-compatible formula string: "[0]+[1]*[5]+[2]*[6]+[3]*[7]+[4]*[8]" --- FDA : Creating and compiling formula --- FDA_Fitter_M...: Parsing option string: --- FDA_Fitter_M...: "!V=False:!H=True:!Formula=(0)+(1)*x0+(2)*x1+(3)*x2+(4)*x3:!ParRanges=(-1,1);(-10,10);(-10,10);(-10,10);(-10,10):!FitMethod=MINUIT:!D=False:!Normalise=False:!VarTransform=None:!VarTransformType=Signal:!NbinsMVAPdf=60:!NsmoothMVAPdf=2:!VerboseLevel=Default:!CreateMVAPdfs=False:!TxtWeightFilesOnly=True:!Converger=None" --- FDA_Fitter_M...: The following options are set: --- FDA_Fitter_M...: - By User: --- FDA_Fitter_M...: <none> --- FDA_Fitter_M...: - Default: --- FDA_Fitter_M...: ErrorLevel: "1" [TMinuit: error level: 0.5=logL fit, 1=chi-squared fit] --- FDA_Fitter_M...: PrintLevel: "-1" [TMinuit: output level: -1=least, 0, +1=all garbage] --- FDA_Fitter_M...: FitStrategy: "2" [TMinuit: fit strategy: 2=best] --- FDA_Fitter_M...: PrintWarnings: "False" [TMinuit: suppress warnings] --- FDA_Fitter_M...: UseImprove: "True" [TMinuit: use IMPROVE] --- FDA_Fitter_M...: UseMinos: "True" [TMinuit: use MINOS] --- FDA_Fitter_M...: SetBatch: "False" [TMinuit: use batch mode] --- FDA_Fitter_M...: MaxCalls: "1000" [TMinuit: approximate maximum number of function calls] --- FDA_Fitter_M...: Tolerance: "0.1" [TMinuit: tolerance to the function value at the minimum] --- FDA_Fitter_M...: <MinuitFitter> Init --- FDA : Preparing evaluation tree... --- FDA : Elapsed time for evaluation of 4 events: [1;31m0.386 sec[0m --- Factory : Test method: MLP --- MLP : Reading weight file: [1;34mweights/TMVAnalysis_MLP.weights.txt[0m --- MLP : Read method with name <MLP> and title <MLP> --- MLP : Classifier was trained with TMVA Version: 3.9.5 --- MLP : Classifier was trained with ROOT Version: 5.20/00 --- MLP : Create VariableTransformation "None" --- MLP : Building Network --- MLP : Initializing weights --- MLP : Forcing weights --- MLP : Preparing evaluation tree... --- MLP : Elapsed time for evaluation of 4 events: [1;31m0.397 sec[0m --- Factory : Test method: SVM_Gauss --- SVM : Reading weight file: [1;34mweights/TMVAnalysis_SVM_Gauss.weights.txt[0m --- SVM : Read method with name <SVM> and title <SVM_Gauss> --- SVM : Classifier was trained with TMVA Version: 3.9.5 --- SVM : Classifier was trained with ROOT Version: 5.20/00 --- SVM : Create VariableTransformation "None" --- SVM : Preparing evaluation tree... --- SVM : Elapsed time for evaluation of 4 events: [1;31m0.366 sec[0m --- Factory : Test method: BDT --- BDT : Reading weight file: [1;34mweights/TMVAnalysis_BDT.weights.txt[0m --- BDT : Read method with name <BDT> and title <BDT> --- BDT : Classifier was trained with TMVA Version: 3.9.5 --- BDT : Classifier was trained with ROOT Version: 5.20/00 --- BDT : Create VariableTransformation "None" --- BDT : Read 400 Decision trees --- BDT : Preparing evaluation tree... --- BDT : Elapsed time for evaluation of 4 events: [1;31m0.445 sec[0m --- Factory : Test method: BDTD --- BDT : Reading weight file: [1;34mweights/TMVAnalysis_BDTD.weights.txt[0m --- BDT : Read method with name <BDT> and title <BDTD> --- BDT : Classifier was trained with TMVA Version: 3.9.5 --- BDT : Classifier was trained with ROOT Version: 5.20/00 --- BDT : Create VariableTransformation "Decorrelate" --- BDT : Read 400 Decision trees --- BDT : Preparing evaluation tree... --- BDT : Elapsed time for evaluation of 4 events: [1;31m0.403 sec[0m --- Factory : Test method: RuleFit --- RuleFit : Reading weight file: [1;34mweights/TMVAnalysis_RuleFit.weights.txt[0m --- RuleFit : Read method with name <RuleFit> and title <RuleFit> --- RuleFit : Classifier was trained with TMVA Version: 3.9.5 --- RuleFit : Classifier was trained with ROOT Version: 5.20/00 --- RuleFit : Create VariableTransformation "None" --- RuleFit : Preparing evaluation tree... --- RuleFit : Elapsed time for evaluation of 4 events: [1;31m0.416 sec[0m --- Factory : Evaluating all classifiers... --- Factory : Evaluate classifier: CutsGA --- Factory : Evaluate classifier: Likelihood --- Likelihood : Loop over test events and fill histograms with classifier response ... --- Factory : Evaluate classifier: LikelihoodPCA --- Likelihood : Loop over test events and fill histograms with classifier response ... --- Factory : Evaluate classifier: PDERS --- PDERS : Loop over test events and fill histograms with classifier response ... --- Factory : Evaluate classifier: KNN --- KNN : Loop over test events and fill histograms with classifier response ... --- Factory : Evaluate classifier: HMatrix --- HMatrix : Loop over test events and fill histograms with classifier response ... --- Factory : Evaluate classifier: Fisher --- Fisher : Loop over test events and fill histograms with classifier response ... --- Fisher : Also filling probability and rarity histograms (on request) ... --- Factory : Evaluate classifier: FDA_MT --- FDA : Loop over test events and fill histograms with classifier response ... --- Factory : Evaluate classifier: MLP --- MLP : Loop over test events and fill histograms with classifier response ... --- Factory : Evaluate classifier: SVM_Gauss --- SVM : Loop over test events and fill histograms with classifier response ... --- Factory : Evaluate classifier: BDT --- BDT : Loop over test events and fill histograms with classifier response ... --- Factory : Evaluate classifier: BDTD --- BDT : Loop over test events and fill histograms with classifier response ... --- Factory : Evaluate classifier: RuleFit --- RuleFit : Loop over test events and fill histograms with classifier response ... --- Factory : --- Factory : Inter-MVA correlation matrix (signal): --- Factory : ------------------------------------------------------------------------------------------------------------------------- --- Factory : Likelihood LikelihoodPCA PDERS KNN HMatrix Fisher FDA_MT MLP SVM_Gauss BDT BDTD RuleFit --- Factory : Likelihood: +1.000 +1.000 -1.000 -1.000 -1.000 -1.000 -1.000 -1.000 -1.000 -1.000 -1.000 +1.000 --- Factory : LikelihoodPCA: +1.000 +1.000 -1.000 -1.000 -1.000 -1.000 -1.000 -1.000 -1.000 -1.000 -1.000 +1.000 --- Factory : PDERS: -1.000 -1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 -1.000 --- Factory : KNN: -1.000 -1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 -1.000 --- Factory : HMatrix: -1.000 -1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 -1.000 --- Factory : Fisher: -1.000 -1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 -1.000 --- Factory : FDA_MT: -1.000 -1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 -1.000 --- Factory : MLP: -1.000 -1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 -1.000 --- Factory : SVM_Gauss: -1.000 -1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 -1.000 --- Factory : BDT: -1.000 -1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 -1.000 --- Factory : BDTD: -1.000 -1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 -1.000 --- Factory : RuleFit: +1.000 +1.000 -1.000 -1.000 -1.000 -1.000 -1.000 -1.000 -1.000 -1.000 -1.000 +1.000 --- Factory : ------------------------------------------------------------------------------------------------------------------------- --- Factory : --- Factory : Inter-MVA correlation matrix (background): --- Factory : ------------------------------------------------------------------------------------------------------------------------- --- Factory : Likelihood LikelihoodPCA PDERS KNN HMatrix Fisher FDA_MT MLP SVM_Gauss BDT BDTD RuleFit --- Factory : Likelihood: +1.000 +1.000 +1.000 +1.000 -1.000 +1.000 +1.000 +1.000 -1.000 +1.000 +1.000 +1.000 --- Factory : LikelihoodPCA: +1.000 +1.000 +1.000 +1.000 -1.000 +1.000 +1.000 +1.000 -1.000 +1.000 +1.000 +1.000 --- Factory : PDERS: +1.000 +1.000 +1.000 +1.000 -1.000 +1.000 +1.000 +1.000 -1.000 +1.000 +1.000 +1.000 --- Factory : KNN: +1.000 +1.000 +1.000 +1.000 -1.000 +1.000 +1.000 +1.000 -1.000 +1.000 +1.000 +1.000 --- Factory : HMatrix: -1.000 -1.000 -1.000 -1.000 +1.000 -1.000 -1.000 -1.000 +1.000 -1.000 -1.000 -1.000 --- Factory : Fisher: +1.000 +1.000 +1.000 +1.000 -1.000 +1.000 +1.000 +1.000 -1.000 +1.000 +1.000 +1.000 --- Factory : FDA_MT: +1.000 +1.000 +1.000 +1.000 -1.000 +1.000 +1.000 +1.000 -1.000 +1.000 +1.000 +1.000 --- Factory : MLP: +1.000 +1.000 +1.000 +1.000 -1.000 +1.000 +1.000 +1.000 -1.000 +1.000 +1.000 +1.000 --- Factory : SVM_Gauss: -1.000 -1.000 -1.000 -1.000 +1.000 -1.000 -1.000 -1.000 +1.000 -1.000 -1.000 -1.000 --- Factory : BDT: +1.000 +1.000 +1.000 +1.000 -1.000 +1.000 +1.000 +1.000 -1.000 +1.000 +1.000 +1.000 --- Factory : BDTD: +1.000 +1.000 +1.000 +1.000 -1.000 +1.000 +1.000 +1.000 -1.000 +1.000 +1.000 +1.000 --- Factory : RuleFit: +1.000 +1.000 +1.000 +1.000 -1.000 +1.000 +1.000 +1.000 -1.000 +1.000 +1.000 +1.000 --- Factory : ------------------------------------------------------------------------------------------------------------------------- --- Factory : --- Factory : The following "overlap" matrices contain the fraction of events for which --- Factory : the MVAs 'i' and 'j' have returned conform answers about "signal-likeness" --- Factory : An event is signal-like, if its MVA output exceeds the following value: --- Factory : ----------------------------- --- Factory : Method: Cut value: --- Factory : ----------------------------- --- Factory : Likelihood: +0.620 --- Factory : LikelihoodPCA: +0.600 --- Factory : PDERS: +0.599 --- Factory : KNN: +0.484 --- Factory : HMatrix: +0.016 --- Factory : Fisher: -0.001 --- Factory : FDA_MT: +0.504 --- Factory : MLP: -0.426 --- Factory : SVM_Gauss: +0.593 --- Factory : BDT: -0.064 --- Factory : BDTD: +0.004 --- Factory : RuleFit: -2.297 --- Factory : ----------------------------- --- Factory : which correspond to the working point: eff(signal) = 1 - eff(background) --- Factory : Note: no correlations and overlap with cut method are provided at present --- Factory : --- Factory : Inter-MVA overlap matrix (signal): --- Factory : ------------------------------------------------------------------------------------------------------------------------- --- Factory : Likelihood LikelihoodPCA PDERS KNN HMatrix Fisher FDA_MT MLP SVM_Gauss BDT BDTD RuleFit --- Factory : Likelihood: +1.000 +0.500 +1.000 +1.000 +0.500 +0.500 +0.500 +0.500 +0.500 +0.500 +0.500 +0.500 --- Factory : LikelihoodPCA: +0.500 +1.000 +0.500 +0.500 +0.000 +0.000 +0.000 +0.000 +0.000 +0.000 +0.000 +1.000 --- Factory : PDERS: +1.000 +0.500 +1.000 +1.000 +0.500 +0.500 +0.500 +0.500 +0.500 +0.500 +0.500 +0.500 --- Factory : KNN: +1.000 +0.500 +1.000 +1.000 +0.500 +0.500 +0.500 +0.500 +0.500 +0.500 +0.500 +0.500 --- Factory : HMatrix: +0.500 +0.000 +0.500 +0.500 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +0.000 --- Factory : Fisher: +0.500 +0.000 +0.500 +0.500 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +0.000 --- Factory : FDA_MT: +0.500 +0.000 +0.500 +0.500 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +0.000 --- Factory : MLP: +0.500 +0.000 +0.500 +0.500 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +0.000 --- Factory : SVM_Gauss: +0.500 +0.000 +0.500 +0.500 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +0.000 --- Factory : BDT: +0.500 +0.000 +0.500 +0.500 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +0.000 --- Factory : BDTD: +0.500 +0.000 +0.500 +0.500 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +1.000 +0.000 --- Factory : RuleFit: +0.500 +1.000 +0.500 +0.500 +0.000 +0.000 +0.000 +0.000 +0.000 +0.000 +0.000 +1.000 --- Factory : ------------------------------------------------------------------------------------------------------------------------- --- Factory : --- Factory : Inter-MVA overlap matrix (background): --- Factory : ------------------------------------------------------------------------------------------------------------------------- --- Factory : Likelihood LikelihoodPCA PDERS KNN HMatrix Fisher FDA_MT MLP SVM_Gauss BDT BDTD RuleFit --- Factory : Likelihood: +1.000 +1.000 +1.000 +1.000 +1.000 +0.500 +0.500 +1.000 +0.500 +1.000 +1.000 +1.000 --- Factory : LikelihoodPCA: +1.000 +1.000 +1.000 +1.000 +1.000 +0.500 +0.500 +1.000 +0.500 +1.000 +1.000 +1.000 --- Factory : PDERS: +1.000 +1.000 +1.000 +1.000 +1.000 +0.500 +0.500 +1.000 +0.500 +1.000 +1.000 +1.000 --- Factory : KNN: +1.000 +1.000 +1.000 +1.000 +1.000 +0.500 +0.500 +1.000 +0.500 +1.000 +1.000 +1.000 --- Factory : HMatrix: +1.000 +1.000 +1.000 +1.000 +1.000 +0.500 +0.500 +1.000 +0.500 +1.000 +1.000 +1.000 --- Factory : Fisher: +0.500 +0.500 +0.500 +0.500 +0.500 +1.000 +1.000 +0.500 +0.000 +0.500 +0.500 +0.500 --- Factory : FDA_MT: +0.500 +0.500 +0.500 +0.500 +0.500 +1.000 +1.000 +0.500 +0.000 +0.500 +0.500 +0.500 --- Factory : MLP: +1.000 +1.000 +1.000 +1.000 +1.000 +0.500 +0.500 +1.000 +0.500 +1.000 +1.000 +1.000 --- Factory : SVM_Gauss: +0.500 +0.500 +0.500 +0.500 +0.500 +0.000 +0.000 +0.500 +1.000 +0.500 +0.500 +0.500 --- Factory : BDT: +1.000 +1.000 +1.000 +1.000 +1.000 +0.500 +0.500 +1.000 +0.500 +1.000 +1.000 +1.000 --- Factory : BDTD: +1.000 +1.000 +1.000 +1.000 +1.000 +0.500 +0.500 +1.000 +0.500 +1.000 +1.000 +1.000 --- Factory : RuleFit: +1.000 +1.000 +1.000 +1.000 +1.000 +0.500 +0.500 +1.000 +0.500 +1.000 +1.000 +1.000 --- Factory : ------------------------------------------------------------------------------------------------------------------------- --- Factory : --- Factory : Evaluation results ranked by best signal efficiency and purity (area) --- Factory : ----------------------------------------------------------------------------- --- Factory : MVA Signal efficiency at bkg eff. (error): | Sepa- Signifi- --- Factory : Methods: @B=0.01 @B=0.10 @B=0.30 Area | ration: cance: --- Factory : ----------------------------------------------------------------------------- --- Factory : Likelihood : 1.000(14) 1.000(14) 1.000(14) 1.000 | 0.912 1.332 --- Factory : LikelihoodPCA : 1.000(14) 1.000(14) 1.000(14) 1.000 | 0.905 1.161 --- Factory : PDERS : 1.000(14) 1.000(14) 1.000(14) 1.000 | 0.912 2.675 --- Factory : KNN : 1.000(14) 1.000(14) 1.000(14) 1.000 | 0.958 15.162 --- Factory : HMatrix : 1.000(14) 1.000(14) 1.000(14) 1.000 | 0.903 3.425 --- Factory : MLP : 1.000(14) 1.000(14) 1.000(14) 1.000 | 0.912 1.270 --- Factory : BDTD : 1.000(14) 1.000(14) 1.000(14) 1.000 | 0.912 4.728 --- Factory : RuleFit : 1.000(14) 1.000(14) 1.000(14) 1.000 | 0.920 1.139 --- Factory : CutsGA : 0.685(328) 0.686(328) 0.689(327) 0.855 | 0.000 0.000 --- Factory : Fisher : 0.495(321) 0.496(321) 0.499(321) 0.626 | 0.912 0.948 --- Factory : FDA_MT : 0.495(321) 0.496(321) 0.499(321) 0.626 | 0.912 1.035 --- Factory : SVM_Gauss : 0.495(321) 0.496(321) 0.498(321) 0.500 | 0.908 0.059 --- Factory : BDT : 0.495(321) 0.496(321) 0.498(321) 0.500 | 0.892 0.544 --- Factory : ----------------------------------------------------------------------------- --- Factory : --- Factory : Testing efficiency compared to training efficiency (overtraining check) --- Factory : ----------------------------------------------------------------------------- --- Factory : MVA Signal efficiency: from test sample (from traing sample) --- Factory : Methods: @B=0.01 @B=0.10 @B=0.30 --- Factory : ----------------------------------------------------------------------------- --- Factory : Likelihood : 1.000 (0.497) 1.000 (1.000) 1.000 (1.000) --- Factory : LikelihoodPCA : 1.000 (0.495) 1.000 (0.499) 1.000 (1.000) --- Factory : PDERS : 1.000 (0.497) 1.000 (1.000) 1.000 (1.000) --- Factory : KNN : 1.000 (1.000) 1.000 (1.000) 1.000 (1.000) --- Factory : HMatrix : 1.000 (0.498) 1.000 (1.000) 1.000 (1.000) --- Factory : MLP : 1.000 (0.495) 1.000 (0.499) 1.000 (1.000) --- Factory : BDTD : 1.000 (0.500) 1.000 (1.000) 1.000 (1.000) --- Factory : RuleFit : 1.000 (0.497) 1.000 (1.000) 1.000 (1.000) --- Factory : CutsGA : 0.685 (0.111) 0.686 (0.492) 0.689 (0.702) --- Factory : Fisher : 0.495 (0.495) 0.496 (0.498) 0.499 (0.503) --- Factory : FDA_MT : 0.495 (0.495) 0.496 (0.498) 0.499 (0.503) --- Factory : SVM_Gauss : 0.495 (0.495) 0.496 (0.496) 0.498 (0.499) --- Factory : BDT : 0.495 (0.495) 0.496 (0.497) 0.498 (0.502) --- Factory : ----------------------------------------------------------------------------- --- Factory : --- Factory : Write Test Tree 'TestTree' to file === wrote root file TMVAout.root === TMVAnalysis is done! --- Launch TMVA GUI to view input file: TMVAout.root --- Reading keys ... === Note: inactive buttons indicate that the corresponding classifiers were not trained ===-- PatRyan - 14 Nov 2008

Edit | Attach | Print version | History: r1 | Backlinks | View wiki text | Edit wiki text | More topic actions

Topic revision: r1 - 14 Nov 2008, PatRyan

Copyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.

Ideas, requests, problems regarding Foswiki? Send feedback

Ideas, requests, problems regarding Foswiki? Send feedback