Logo ROOT   6.14/05
Reference Guide
TMVARegression.C File Reference
Tutorials » TMVA tutorials

Detailed Description

View in nbviewer Open in SWAN This macro provides examples for the training and testing of the TMVA classifiers.

As input data is used a toy-MC sample consisting of four Gaussian-distributed and linearly correlated input variables.

The methods to be used can be switched on and off by means of booleans, or via the prompt command, for example: 

root -l TMVARegression.C\(\"LD,MLP\"\)

(note that the backslashes are mandatory) If no method given, a default set is used.

The output file "TMVAReg.root" can be analysed with the use of dedicated macros (simply say: root -l <macro.C>), which can be conveniently invoked through a GUI that will appear at the end of the run of this macro.

0.060357093811
25.8791851997
Processing /mnt/build/workspace/root-makedoc-v614/rootspi/rdoc/src/v6-14-00-patches/tutorials/tmva/TMVARegression.C...
==> Start TMVARegression
--- TMVARegression : Using input file: ./files/tmva_reg_example.root
DataSetInfo : [dataset] : Added class "Regression"
: Add Tree TreeR of type Regression with 10000 events
: Dataset[dataset] : Class index : 0 name : Regression
Factory : Booking method: PDEFoam
:
DataSetFactory : [dataset] : Number of events in input trees
:
: Number of training and testing events
: ---------------------------------------------------------------------------
: Regression -- training events : 1000
: Regression -- testing events : 9000
: Regression -- training and testing events: 10000
:
DataSetInfo : Correlation matrix (Regression):
: ------------------------
: var1 var2
: var1: +1.000 -0.018
: var2: -0.018 +1.000
: ------------------------
DataSetFactory : [dataset] :
:
Factory : Booking method: KNN
:
Factory : Booking method: LD
:
Factory : Booking method: DNN_CPU
:
: Parsing option string:
: ... "!H:V:ErrorStrategy=SUMOFSQUARES:VarTransform=G:WeightInitialization=XAVIERUNIFORM:Architecture=CPU:Layout=TANH|50,Layout=TANH|50,Layout=TANH|50,LINEAR:TrainingStrategy=LearningRate=1e-2,Momentum=0.5,Repetitions=1,ConvergenceSteps=20,BatchSize=50,TestRepetitions=10,WeightDecay=0.01,Regularization=NONE,DropConfig=0.2+0.2+0.2+0.,DropRepetitions=2|LearningRate=1e-3,Momentum=0.9,Repetitions=1,ConvergenceSteps=20,BatchSize=50,TestRepetitions=5,WeightDecay=0.01,Regularization=L2,DropConfig=0.1+0.1+0.1,DropRepetitions=1|LearningRate=1e-4,Momentum=0.3,Repetitions=1,ConvergenceSteps=10,BatchSize=50,TestRepetitions=5,WeightDecay=0.01,Regularization=NONE"
: The following options are set:
: - By User:
: <none>
: - Default:
: Boost_num: "0" [Number of times the classifier will be boosted]
: Parsing option string:
: ... "!H:V:ErrorStrategy=SUMOFSQUARES:VarTransform=G:WeightInitialization=XAVIERUNIFORM:Architecture=CPU:Layout=TANH|50,Layout=TANH|50,Layout=TANH|50,LINEAR:TrainingStrategy=LearningRate=1e-2,Momentum=0.5,Repetitions=1,ConvergenceSteps=20,BatchSize=50,TestRepetitions=10,WeightDecay=0.01,Regularization=NONE,DropConfig=0.2+0.2+0.2+0.,DropRepetitions=2|LearningRate=1e-3,Momentum=0.9,Repetitions=1,ConvergenceSteps=20,BatchSize=50,TestRepetitions=5,WeightDecay=0.01,Regularization=L2,DropConfig=0.1+0.1+0.1,DropRepetitions=1|LearningRate=1e-4,Momentum=0.3,Repetitions=1,ConvergenceSteps=10,BatchSize=50,TestRepetitions=5,WeightDecay=0.01,Regularization=NONE"
: The following options are set:
: - By User:
: V: "True" [Verbose output (short form of "VerbosityLevel" below - overrides the latter one)]
: VarTransform: "G" [List of variable transformations performed before training, e.g., "D_Background,P_Signal,G,N_AllClasses" for: "Decorrelation, PCA-transformation, Gaussianisation, Normalisation, each for the given class of events ('AllClasses' denotes all events of all classes, if no class indication is given, 'All' is assumed)"]
: H: "False" [Print method-specific help message]
: Layout: "TANH|50,Layout=TANH|50,Layout=TANH|50,LINEAR" [Layout of the network.]
: ErrorStrategy: "SUMOFSQUARES" [Loss function: Mean squared error (regression) or cross entropy (binary classification).]
: WeightInitialization: "XAVIERUNIFORM" [Weight initialization strategy]
: Architecture: "CPU" [Which architecture to perform the training on.]
: TrainingStrategy: "LearningRate=1e-2,Momentum=0.5,Repetitions=1,ConvergenceSteps=20,BatchSize=50,TestRepetitions=10,WeightDecay=0.01,Regularization=NONE,DropConfig=0.2+0.2+0.2+0.,DropRepetitions=2|LearningRate=1e-3,Momentum=0.9,Repetitions=1,ConvergenceSteps=20,BatchSize=50,TestRepetitions=5,WeightDecay=0.01,Regularization=L2,DropConfig=0.1+0.1+0.1,DropRepetitions=1|LearningRate=1e-4,Momentum=0.3,Repetitions=1,ConvergenceSteps=10,BatchSize=50,TestRepetitions=5,WeightDecay=0.01,Regularization=NONE" [Defines the training strategies.]
: - Default:
: VerbosityLevel: "Default" [Verbosity level]
: CreateMVAPdfs: "False" [Create PDFs for classifier outputs (signal and background)]
: IgnoreNegWeightsInTraining: "False" [Events with negative weights are ignored in the training (but are included for testing and performance evaluation)]
: ValidationSize: "20%" [Part of the training data to use for validation. Specify as 0.2 or 20% to use a fifth of the data set as validation set. Specify as 100 to use exactly 100 events. (Default: 20%)]
DNN_CPU : [dataset] : Create Transformation "G" with events from all classes.
:
: Transformation, Variable selection :
: Input : variable 'var1' <---> Output : variable 'var1'
: Input : variable 'var2' <---> Output : variable 'var2'
: Preparing the Gaussian transformation...
TFHandler_DNN_CPU : Variable Mean RMS [ Min Max ]
: -----------------------------------------------------------
: var1: 0.013183 1.0272 [ -3.3668 5.7307 ]
: var2: 0.0071633 1.0351 [ -4.2630 5.7307 ]
: fvalue: 164.96 82.203 [ 1.7144 391.23 ]
: -----------------------------------------------------------
Parsed Training DNN string LearningRate=1e-2,Momentum=0.5,Repetitions=1,ConvergenceSteps=20,BatchSize=50,TestRepetitions=10,WeightDecay=0.01,Regularization=NONE,DropConfig=0.2+0.2+0.2+0.,DropRepetitions=2|LearningRate=1e-3,Momentum=0.9,Repetitions=1,ConvergenceSteps=20,BatchSize=50,TestRepetitions=5,WeightDecay=0.01,Regularization=L2,DropConfig=0.1+0.1+0.1,DropRepetitions=1|LearningRate=1e-4,Momentum=0.3,Repetitions=1,ConvergenceSteps=10,BatchSize=50,TestRepetitions=5,WeightDecay=0.01,Regularization=NONE
STring has size 3
Factory : Booking method: BDTG
:
<WARNING> : Value for option maxdepth was previously set to 3
: the option *InverseBoostNegWeights* does not exist for BoostType=Grad --> change
: to new default for GradBoost *Pray*
Factory : Train all methods
Factory : [dataset] : Create Transformation "I" with events from all classes.
:
: Transformation, Variable selection :
: Input : variable 'var1' <---> Output : variable 'var1'
: Input : variable 'var2' <---> Output : variable 'var2'
TFHandler_Factory : Variable Mean RMS [ Min Max ]
: -----------------------------------------------------------
: var1: 3.4138 1.1963 [ 0.0026062 4.9957 ]
: var2: 2.4356 1.4134 [ 0.0092062 4.9990 ]
: fvalue: 164.96 82.203 [ 1.7144 391.23 ]
: -----------------------------------------------------------
: Ranking input variables (method unspecific)...
IdTransformation : Ranking result (top variable is best ranked)
: --------------------------------------------
: Rank : Variable : |Correlation with target|
: --------------------------------------------
: 1 : var2 : 7.419e-01
: 2 : var1 : 5.996e-01
: --------------------------------------------
IdTransformation : Ranking result (top variable is best ranked)
: -------------------------------------
: Rank : Variable : Mutual information
: -------------------------------------
: 1 : var2 : 2.029e+00
: 2 : var1 : 1.950e+00
: -------------------------------------
IdTransformation : Ranking result (top variable is best ranked)
: ------------------------------------
: Rank : Variable : Correlation Ratio
: ------------------------------------
: 1 : var1 : 6.538e+00
: 2 : var2 : 2.460e+00
: ------------------------------------
IdTransformation : Ranking result (top variable is best ranked)
: ----------------------------------------
: Rank : Variable : Correlation Ratio (T)
: ----------------------------------------
: 1 : var2 : 9.156e-01
: 2 : var1 : 2.981e-01
: ----------------------------------------
Factory : Train method: PDEFoam for Regression
:
: Build mono target regression foam
: Elapsed time: 0.646 sec
: Elapsed time for training with 1000 events: 0.653 sec
: Dataset[dataset] : Create results for training
: Dataset[dataset] : Evaluation of PDEFoam on training sample
: Dataset[dataset] : Elapsed time for evaluation of 1000 events: 0.0071 sec
: Create variable histograms
: Create regression target histograms
: Create regression average deviation
: Results created
: Creating xml weight file: dataset/weights/TMVARegression_PDEFoam.weights.xml
: writing foam MonoTargetRegressionFoam to file
: Foams written to file: dataset/weights/TMVARegression_PDEFoam.weights_foams.root
Factory : Training finished
:
Factory : Train method: KNN for Regression
:
KNN : <Train> start...
: Reading 1000 events
: Number of signal events 1000
: Number of background events 0
: Creating kd-tree with 1000 events
: Computing scale factor for 1d distributions: (ifrac, bottom, top) = (80%, 10%, 90%)
ModulekNN : Optimizing tree for 2 variables with 1000 values
: <Fill> Class 1 has 1000 events
: Elapsed time for training with 1000 events: 0.00153 sec
: Dataset[dataset] : Create results for training
: Dataset[dataset] : Evaluation of KNN on training sample
: Dataset[dataset] : Elapsed time for evaluation of 1000 events: 0.0135 sec
: Create variable histograms
: Create regression target histograms
: Create regression average deviation
: Results created
: Creating xml weight file: dataset/weights/TMVARegression_KNN.weights.xml
Factory : Training finished
:
Factory : Train method: LD for Regression
:
LD : Results for LD coefficients:
: -----------------------
: Variable: Coefficient:
: -----------------------
: var1: +42.104
: var2: +44.607
: (offset): -87.420
: -----------------------
: Elapsed time for training with 1000 events: 0.000373 sec
: Dataset[dataset] : Create results for training
: Dataset[dataset] : Evaluation of LD on training sample
: Dataset[dataset] : Elapsed time for evaluation of 1000 events: 0.00231 sec
: Create variable histograms
: Create regression target histograms
: Create regression average deviation
: Results created
: Creating xml weight file: dataset/weights/TMVARegression_LD.weights.xml
Factory : Training finished
:
Factory : Train method: DNN_CPU for Regression
:
TFHandler_DNN_CPU : Variable Mean RMS [ Min Max ]
: -----------------------------------------------------------
: var1: 0.013183 1.0272 [ -3.3668 5.7307 ]
: var2: 0.0071633 1.0351 [ -4.2630 5.7307 ]
: fvalue: 164.96 82.203 [ 1.7144 391.23 ]
: -----------------------------------------------------------
: Start of neural network training on CPU.
:
: Training phase 1 of 3:
: Epoch | Train Err. Test Err. GFLOP/s Conv. Steps
: --------------------------------------------------------------
: 10 | 1223.16 1186.64 1.0245 0
: 20 | 1913.8 1799.36 1.1141 10
: 30 | 2989.46 3125.5 1.19686 20
:
: Training phase 2 of 3:
: Epoch | Train Err. Test Err. GFLOP/s Conv. Steps
: --------------------------------------------------------------
: 5 | 7766.39 1639.33 1.20215 0
: 10 | 6828.2 1138.18 1.17525 0
: 15 | 6750.87 1258.19 1.13374 5
: 20 | 6638.98 1359.33 1.13497 10
: 25 | 6426.33 1084 1.12594 0
: 30 | 6358.57 1128.37 1.1136 5
: 35 | 6601.17 1413.66 1.08626 10
: 40 | 6383.97 1606.09 1.12923 15
: 45 | 6674.44 1582.05 1.12947 20
:
: Training phase 3 of 3:
: Epoch | Train Err. Test Err. GFLOP/s Conv. Steps
: --------------------------------------------------------------
: 5 | 1567.14 1408.25 1.57317 0
: 10 | 1524.59 1378.28 1.61924 0
: 15 | 1467.95 1334.19 1.59934 0
: 20 | 1441.84 1322.22 1.63546 0
: 25 | 1428.85 1311.5 1.6441 0
: 30 | 1416.92 1301.79 1.66205 0
: 35 | 1406.14 1293.01 1.64024 0
: 40 | 1396 1283.43 1.61055 0
: 45 | 1386.62 1276 1.64586 0
: 50 | 1377.8 1269.35 1.66474 0
: 55 | 1369.42 1263.2 1.68068 0
: 60 | 1361.67 1258.02 1.65397 0
: 65 | 1353.56 1253.52 1.64161 0
: 70 | 1346.7 1248.81 1.64646 0
: 75 | 1340.95 1245 1.6497 0
: 80 | 1335.81 1241.54 1.55391 0
: 85 | 1330.99 1237.98 1.50878 0
: 90 | 1326.55 1235.22 1.59568 0
: 95 | 1322.52 1232.47 1.58413 0
: 100 | 1318.68 1230.09 1.58085 0
: 105 | 1315.12 1227.62 1.60077 0
: 110 | 1311.76 1225.81 1.61769 0
: 115 | 1308.71 1223.71 1.63288 0
: 120 | 1305.77 1222.09 1.66172 0
: 125 | 1302.99 1220.67 1.63651 0
: 130 | 1300.39 1219.44 1.68469 0
: 135 | 1297.78 1217.81 1.63767 0
: 140 | 1295.39 1216.76 1.64437 5
: 145 | 1293.29 1215.88 1.57922 0
: 150 | 1291.21 1214.68 1.63066 5
: 155 | 1289.33 1213.91 1.67057 0
: 160 | 1287.47 1212.89 1.65315 5
: 165 | 1285.79 1212.25 1.71419 0
: 170 | 1284.07 1211.74 1.65769 5
: 175 | 1282.55 1211 1.62855 0
: 180 | 1280.92 1210.58 1.6441 5
: 185 | 1279.37 1210.09 1.66901 10
:
: Elapsed time for training with 1000 events: 4.85 sec
: Dataset[dataset] : Create results for training
: Dataset[dataset] : Evaluation of DNN_CPU on training sample
: Dataset[dataset] : Elapsed time for evaluation of 1000 events: 0.0267 sec
: Create variable histograms
: Create regression target histograms
: Create regression average deviation
: Results created
: Creating xml weight file: dataset/weights/TMVARegression_DNN_CPU.weights.xml
Factory : Training finished
:
Factory : Train method: BDTG for Regression
:
: Regression Loss Function: Huber
: Training 2000 Decision Trees ... patience please
: Elapsed time for training with 1000 events: 2.15 sec
: Dataset[dataset] : Create results for training
: Dataset[dataset] : Evaluation of BDTG on training sample
: Dataset[dataset] : Elapsed time for evaluation of 1000 events: 0.377 sec
: Create variable histograms
: Create regression target histograms
: Create regression average deviation
: Results created
: Creating xml weight file: dataset/weights/TMVARegression_BDTG.weights.xml
Factory : Training finished
:
Factory : === Destroy and recreate all methods via weight files for testing ===
:
: Reading weight file: dataset/weights/TMVARegression_PDEFoam.weights.xml
: Read foams from file: dataset/weights/TMVARegression_PDEFoam.weights_foams.root
: Reading weight file: dataset/weights/TMVARegression_KNN.weights.xml
: Creating kd-tree with 1000 events
: Computing scale factor for 1d distributions: (ifrac, bottom, top) = (80%, 10%, 90%)
ModulekNN : Optimizing tree for 2 variables with 1000 values
: <Fill> Class 1 has 1000 events
: Reading weight file: dataset/weights/TMVARegression_LD.weights.xml
: Reading weight file: dataset/weights/TMVARegression_DNN_CPU.weights.xml
: Reading weight file: dataset/weights/TMVARegression_BDTG.weights.xml
Factory : Test all methods
Factory : Test method: PDEFoam for Regression performance
:
: Dataset[dataset] : Create results for testing
: Dataset[dataset] : Evaluation of PDEFoam on testing sample
: Dataset[dataset] : Elapsed time for evaluation of 9000 events: 0.0535 sec
: Create variable histograms
: Create regression target histograms
: Create regression average deviation
: Results created
Factory : Test method: KNN for Regression performance
:
: Dataset[dataset] : Create results for testing
: Dataset[dataset] : Evaluation of KNN on testing sample
: Dataset[dataset] : Elapsed time for evaluation of 9000 events: 0.113 sec
: Create variable histograms
: Create regression target histograms
: Create regression average deviation
: Results created
Factory : Test method: LD for Regression performance
:
: Dataset[dataset] : Create results for testing
: Dataset[dataset] : Evaluation of LD on testing sample
: Dataset[dataset] : Elapsed time for evaluation of 9000 events: 0.00821 sec
: Create variable histograms
: Create regression target histograms
: Create regression average deviation
: Results created
Factory : Test method: DNN_CPU for Regression performance
:
: Dataset[dataset] : Create results for testing
: Dataset[dataset] : Evaluation of DNN_CPU on testing sample
: Dataset[dataset] : Elapsed time for evaluation of 9000 events: 0.25 sec
: Create variable histograms
: Create regression target histograms
: Create regression average deviation
: Results created
Factory : Test method: BDTG for Regression performance
:
: Dataset[dataset] : Create results for testing
: Dataset[dataset] : Evaluation of BDTG on testing sample
: Dataset[dataset] : Elapsed time for evaluation of 9000 events: 1.82 sec
: Create variable histograms
: Create regression target histograms
: Create regression average deviation
: Results created
Factory : Evaluate all methods
: Evaluate regression method: PDEFoam
: TestRegression (testing)
: Calculate regression for all events
: Elapsed time for evaluation of 9000 events: 0.047 sec
: TestRegression (training)
: Calculate regression for all events
: Elapsed time for evaluation of 1000 events: 0.00592 sec
TFHandler_PDEFoam : Variable Mean RMS [ Min Max ]
: -----------------------------------------------------------
: var1: 3.3309 1.1858 [ 0.00020069 5.0000 ]
: var2: 2.4914 1.4393 [ 0.00071490 5.0000 ]
: fvalue: 164.02 83.932 [ 1.6186 394.84 ]
: -----------------------------------------------------------
: Evaluate regression method: KNN
: TestRegression (testing)
: Calculate regression for all events
: Elapsed time for evaluation of 9000 events: 0.107 sec
: TestRegression (training)
: Calculate regression for all events
: Elapsed time for evaluation of 1000 events: 0.0115 sec
TFHandler_KNN : Variable Mean RMS [ Min Max ]
: -----------------------------------------------------------
: var1: 3.3309 1.1858 [ 0.00020069 5.0000 ]
: var2: 2.4914 1.4393 [ 0.00071490 5.0000 ]
: fvalue: 164.02 83.932 [ 1.6186 394.84 ]
: -----------------------------------------------------------
: Evaluate regression method: LD
: TestRegression (testing)
: Calculate regression for all events
: Elapsed time for evaluation of 9000 events: 0.00771 sec
: TestRegression (training)
: Calculate regression for all events
: Elapsed time for evaluation of 1000 events: 0.00122 sec
TFHandler_LD : Variable Mean RMS [ Min Max ]
: -----------------------------------------------------------
: var1: 3.3309 1.1858 [ 0.00020069 5.0000 ]
: var2: 2.4914 1.4393 [ 0.00071490 5.0000 ]
: fvalue: 164.02 83.932 [ 1.6186 394.84 ]
: -----------------------------------------------------------
: Evaluate regression method: DNN_CPU
: TestRegression (testing)
: Calculate regression for all events
: Elapsed time for evaluation of 9000 events: 0.231 sec
: TestRegression (training)
: Calculate regression for all events
: Elapsed time for evaluation of 1000 events: 0.0287 sec
TFHandler_DNN_CPU : Variable Mean RMS [ Min Max ]
: -----------------------------------------------------------
: var1: -0.062720 1.0031 [ -3.3827 5.7307 ]
: var2: 0.031261 1.0685 [ -5.7307 5.7307 ]
: fvalue: 164.02 83.932 [ 1.6186 394.84 ]
: -----------------------------------------------------------
TFHandler_DNN_CPU : Variable Mean RMS [ Min Max ]
: -----------------------------------------------------------
: var1: -0.062720 1.0031 [ -3.3827 5.7307 ]
: var2: 0.031261 1.0685 [ -5.7307 5.7307 ]
: fvalue: 164.02 83.932 [ 1.6186 394.84 ]
: -----------------------------------------------------------
: Evaluate regression method: BDTG
: TestRegression (testing)
: Calculate regression for all events
: Elapsed time for evaluation of 9000 events: 1.96 sec
: TestRegression (training)
: Calculate regression for all events
: Elapsed time for evaluation of 1000 events: 0.197 sec
TFHandler_BDTG : Variable Mean RMS [ Min Max ]
: -----------------------------------------------------------
: var1: 3.3309 1.1858 [ 0.00020069 5.0000 ]
: var2: 2.4914 1.4393 [ 0.00071490 5.0000 ]
: fvalue: 164.02 83.932 [ 1.6186 394.84 ]
: -----------------------------------------------------------
:
: Evaluation results ranked by smallest RMS on test sample:
: ("Bias" quotes the mean deviation of the regression from true target.
: "MutInf" is the "Mutual Information" between regression and target.
: Indicated by "_T" are the corresponding "truncated" quantities ob-
: tained when removing events deviating more than 2sigma from average.)
: --------------------------------------------------------------------------------------------------
: --------------------------------------------------------------------------------------------------
: dataset KNN : -0.507 0.436 5.77 3.79 | 2.871 2.903
: dataset PDEFoam : -0.831 -0.645 9.90 8.12 | 2.245 2.327
: dataset LD : -0.0644 1.63 19.7 17.9 | 1.988 1.981
: dataset DNN_CPU : -1.31 3.92 37.7 30.0 | 1.153 1.166
: dataset BDTG : 0.531 -5.57 82.5 73.9 | 2.307 2.239
: --------------------------------------------------------------------------------------------------
:
: Evaluation results ranked by smallest RMS on training sample:
: (overtraining check)
: --------------------------------------------------------------------------------------------------
: DataSet Name: MVA Method: <Bias> <Bias_T> RMS RMS_T | MutInf MutInf_T
: --------------------------------------------------------------------------------------------------
: dataset KNN : -0.523 0.298 5.55 3.82 | 2.931 2.946
: dataset PDEFoam : 7.41e-07 0.243 7.99 6.37 | 2.489 2.565
: dataset LD : 3.68e-06 1.76 18.9 16.9 | 2.101 2.099
: dataset DNN_CPU : 0.0254 4.80 35.6 28.6 | 1.235 1.247
: dataset BDTG : 0.354 -4.75 80.9 72.3 | 2.369 2.287
: --------------------------------------------------------------------------------------------------
:
Dataset:dataset : Created tree 'TestTree' with 9000 events
:
Dataset:dataset : Created tree 'TrainTree' with 1000 events
:
Factory : Thank you for using TMVA!
: For citation information, please visit: http://tmva.sf.net/citeTMVA.html
==> Wrote root file: TMVAReg.root
==> TMVARegression is done!
#include <cstdlib>
#include <iostream>
#include <map>
#include <string>
#include "TChain.h"
#include "TFile.h"
#include "TTree.h"
#include "TString.h"
#include "TObjString.h"
#include "TSystem.h"
#include "TROOT.h"
#include "TMVA/Tools.h"
#include "TMVA/Factory.h"
#include "TMVA/DataLoader.h"
#include "TMVA/TMVARegGui.h"
using namespace TMVA;
void TMVARegression( TString myMethodList = "" )
{
// The explicit loading of the shared libTMVA is done in TMVAlogon.C, defined in .rootrc
// if you use your private .rootrc, or run from a different directory, please copy the
// corresponding lines from .rootrc
// methods to be processed can be given as an argument; use format:
//
// mylinux~> root -l TMVARegression.C\(\"myMethod1,myMethod2,myMethod3\"\)
//
//---------------------------------------------------------------
// This loads the library
TMVA::Tools::Instance();
// Default MVA methods to be trained + tested
std::map<std::string,int> Use;
// Mutidimensional likelihood and Nearest-Neighbour methods
Use["PDERS"] = 0;
Use["PDEFoam"] = 1;
Use["KNN"] = 1;
//
// Linear Discriminant Analysis
Use["LD"] = 1;
//
// Function Discriminant analysis
Use["FDA_GA"] = 0;
Use["FDA_MC"] = 0;
Use["FDA_MT"] = 0;
Use["FDA_GAMT"] = 0;
//
// Neural Network
Use["MLP"] = 0;
#ifdef R__HAS_TMVACPU
Use["DNN_CPU"] = 1;
#else
Use["DNN_CPU"] = 0;
#endif
//
// Support Vector Machine
Use["SVM"] = 0;
//
// Boosted Decision Trees
Use["BDT"] = 0;
Use["BDTG"] = 1;
// ---------------------------------------------------------------
std::cout << std::endl;
std::cout << "==> Start TMVARegression" << std::endl;
// Select methods (don't look at this code - not of interest)
if (myMethodList != "") {
for (std::map<std::string,int>::iterator it = Use.begin(); it != Use.end(); it++) it->second = 0;
std::vector<TString> mlist = gTools().SplitString( myMethodList, ',' );
for (UInt_t i=0; i<mlist.size(); i++) {
std::string regMethod(mlist[i]);
if (Use.find(regMethod) == Use.end()) {
std::cout << "Method \"" << regMethod << "\" not known in TMVA under this name. Choose among the following:" << std::endl;
for (std::map<std::string,int>::iterator it = Use.begin(); it != Use.end(); it++) std::cout << it->first << " ";
std::cout << std::endl;
return;
}
Use[regMethod] = 1;
}
}
// --------------------------------------------------------------------------------------------------
// Here the preparation phase begins
// Create a new root output file
TString outfileName( "TMVAReg.root" );
TFile* outputFile = TFile::Open( outfileName, "RECREATE" );
// Create the factory object. Later you can choose the methods
// whose performance you'd like to investigate. The factory will
// then run the performance analysis for you.
//
// The first argument is the base of the name of all the
// weightfiles in the directory weight/
//
// The second argument is the output file for the training results
// All TMVA output can be suppressed by removing the "!" (not) in
// front of the "Silent" argument in the option string
TMVA::Factory *factory = new TMVA::Factory( "TMVARegression", outputFile,
"!V:!Silent:Color:DrawProgressBar:AnalysisType=Regression" );
TMVA::DataLoader *dataloader=new TMVA::DataLoader("dataset");
// If you wish to modify default settings
// (please check "src/Config.h" to see all available global options)
//
// (TMVA::gConfig().GetVariablePlotting()).fTimesRMS = 8.0;
// (TMVA::gConfig().GetIONames()).fWeightFileDir = "myWeightDirectory";
// Define the input variables that shall be used for the MVA training
// note that you may also use variable expressions, such as: "3*var1/var2*abs(var3)"
// [all types of expressions that can also be parsed by TTree::Draw( "expression" )]
dataloader->AddVariable( "var1", "Variable 1", "units", 'F' );
dataloader->AddVariable( "var2", "Variable 2", "units", 'F' );
// You can add so-called "Spectator variables", which are not used in the MVA training,
// but will appear in the final "TestTree" produced by TMVA. This TestTree will contain the
// input variables, the response values of all trained MVAs, and the spectator variables
dataloader->AddSpectator( "spec1:=var1*2", "Spectator 1", "units", 'F' );
dataloader->AddSpectator( "spec2:=var1*3", "Spectator 2", "units", 'F' );
// Add the variable carrying the regression target
dataloader->AddTarget( "fvalue" );
// It is also possible to declare additional targets for multi-dimensional regression, ie:
// factory->AddTarget( "fvalue2" );
// BUT: this is currently ONLY implemented for MLP
// Read training and test data (see TMVAClassification for reading ASCII files)
// load the signal and background event samples from ROOT trees
TFile *input(0);
TString fname = "./tmva_reg_example.root";
if (!gSystem->AccessPathName( fname )) {
input = TFile::Open( fname ); // check if file in local directory exists
}
else {
TFile::SetCacheFileDir(".");
input = TFile::Open("http://root.cern.ch/files/tmva_reg_example.root", "CACHEREAD"); // if not: download from ROOT server
}
if (!input) {
std::cout << "ERROR: could not open data file" << std::endl;
exit(1);
}
std::cout << "--- TMVARegression : Using input file: " << input->GetName() << std::endl;
// Register the regression tree
TTree *regTree = (TTree*)input->Get("TreeR");
// global event weights per tree (see below for setting event-wise weights)
Double_t regWeight = 1.0;
// You can add an arbitrary number of regression trees
dataloader->AddRegressionTree( regTree, regWeight );
// This would set individual event weights (the variables defined in the
// expression need to exist in the original TTree)
dataloader->SetWeightExpression( "var1", "Regression" );
// Apply additional cuts on the signal and background samples (can be different)
TCut mycut = ""; // for example: TCut mycut = "abs(var1)<0.5 && abs(var2-0.5)<1";
// tell the DataLoader to use all remaining events in the trees after training for testing:
dataloader->PrepareTrainingAndTestTree( mycut,
"nTrain_Regression=1000:nTest_Regression=0:SplitMode=Random:NormMode=NumEvents:!V" );
//
// dataloader->PrepareTrainingAndTestTree( mycut,
// "nTrain_Regression=0:nTest_Regression=0:SplitMode=Random:NormMode=NumEvents:!V" );
// If no numbers of events are given, half of the events in the tree are used
// for training, and the other half for testing:
//
// dataloader->PrepareTrainingAndTestTree( mycut, "SplitMode=random:!V" );
// Book MVA methods
//
// Please lookup the various method configuration options in the corresponding cxx files, eg:
// src/MethoCuts.cxx, etc, or here: http://tmva.sourceforge.net/optionRef.html
// it is possible to preset ranges in the option string in which the cut optimisation should be done:
// "...:CutRangeMin[2]=-1:CutRangeMax[2]=1"...", where [2] is the third input variable
// PDE - RS method
if (Use["PDERS"])
factory->BookMethod( dataloader, TMVA::Types::kPDERS, "PDERS",
"!H:!V:NormTree=T:VolumeRangeMode=Adaptive:KernelEstimator=Gauss:GaussSigma=0.3:NEventsMin=40:NEventsMax=60:VarTransform=None" );
// And the options strings for the MinMax and RMS methods, respectively:
//
// "!H:!V:VolumeRangeMode=MinMax:DeltaFrac=0.2:KernelEstimator=Gauss:GaussSigma=0.3" );
// "!H:!V:VolumeRangeMode=RMS:DeltaFrac=3:KernelEstimator=Gauss:GaussSigma=0.3" );
if (Use["PDEFoam"])
factory->BookMethod( dataloader, TMVA::Types::kPDEFoam, "PDEFoam",
"!H:!V:MultiTargetRegression=F:TargetSelection=Mpv:TailCut=0.001:VolFrac=0.0666:nActiveCells=500:nSampl=2000:nBin=5:Compress=T:Kernel=None:Nmin=10:VarTransform=None" );
// K-Nearest Neighbour classifier (KNN)
if (Use["KNN"])
factory->BookMethod( dataloader, TMVA::Types::kKNN, "KNN",
"nkNN=20:ScaleFrac=0.8:SigmaFact=1.0:Kernel=Gaus:UseKernel=F:UseWeight=T:!Trim" );
// Linear discriminant
if (Use["LD"])
factory->BookMethod( dataloader, TMVA::Types::kLD, "LD",
"!H:!V:VarTransform=None" );
// Function discrimination analysis (FDA) -- test of various fitters - the recommended one is Minuit (or GA or SA)
if (Use["FDA_MC"])
factory->BookMethod( dataloader, TMVA::Types::kFDA, "FDA_MC",
"!H:!V:Formula=(0)+(1)*x0+(2)*x1:ParRanges=(-100,100);(-100,100);(-100,100):FitMethod=MC:SampleSize=100000:Sigma=0.1:VarTransform=D" );
if (Use["FDA_GA"]) // can also use Simulated Annealing (SA) algorithm (see Cuts_SA options) .. the formula of this example is good for parabolas
factory->BookMethod( dataloader, TMVA::Types::kFDA, "FDA_GA",
"!H:!V:Formula=(0)+(1)*x0+(2)*x1:ParRanges=(-100,100);(-100,100);(-100,100):FitMethod=GA:PopSize=100:Cycles=3:Steps=30:Trim=True:SaveBestGen=1:VarTransform=Norm" );
if (Use["FDA_MT"])
factory->BookMethod( dataloader, TMVA::Types::kFDA, "FDA_MT",
"!H:!V:Formula=(0)+(1)*x0+(2)*x1:ParRanges=(-100,100);(-100,100);(-100,100);(-10,10):FitMethod=MINUIT:ErrorLevel=1:PrintLevel=-1:FitStrategy=2:UseImprove:UseMinos:SetBatch" );
if (Use["FDA_GAMT"])
factory->BookMethod( dataloader, TMVA::Types::kFDA, "FDA_GAMT",
"!H:!V:Formula=(0)+(1)*x0+(2)*x1:ParRanges=(-100,100);(-100,100);(-100,100):FitMethod=GA:Converger=MINUIT:ErrorLevel=1:PrintLevel=-1:FitStrategy=0:!UseImprove:!UseMinos:SetBatch:Cycles=1:PopSize=5:Steps=5:Trim" );
// Neural network (MLP)
if (Use["MLP"])
factory->BookMethod( dataloader, TMVA::Types::kMLP, "MLP", "!H:!V:VarTransform=Norm:NeuronType=tanh:NCycles=20000:HiddenLayers=N+20:TestRate=6:TrainingMethod=BFGS:Sampling=0.3:SamplingEpoch=0.8:ConvergenceImprove=1e-6:ConvergenceTests=15:!UseRegulator" );
if (Use["DNN_CPU"]) {
/*
TString layoutString ("Layout=TANH|(N+100)*2,LINEAR");
TString layoutString ("Layout=SOFTSIGN|100,SOFTSIGN|50,SOFTSIGN|20,LINEAR");
TString layoutString ("Layout=RELU|300,RELU|100,RELU|30,RELU|10,LINEAR");
TString layoutString ("Layout=SOFTSIGN|50,SOFTSIGN|30,SOFTSIGN|20,SOFTSIGN|10,LINEAR");
TString layoutString ("Layout=TANH|50,TANH|30,TANH|20,TANH|10,LINEAR");
TString layoutString ("Layout=SOFTSIGN|50,SOFTSIGN|20,LINEAR");
TString layoutString ("Layout=TANH|100,TANH|30,LINEAR");
*/
TString layoutString("Layout=TANH|50,Layout=TANH|50,Layout=TANH|50,LINEAR");
TString training0("LearningRate=1e-2,Momentum=0.5,Repetitions=1,ConvergenceSteps=20,BatchSize=50,"
"TestRepetitions=10,WeightDecay=0.01,Regularization=NONE,DropConfig=0.2+0.2+0.2+0.,"
"DropRepetitions=2");
TString training1("LearningRate=1e-3,Momentum=0.9,Repetitions=1,ConvergenceSteps=20,BatchSize=50,"
"TestRepetitions=5,WeightDecay=0.01,Regularization=L2,DropConfig=0.1+0.1+0.1,DropRepetitions="
"1");
TString training2("LearningRate=1e-4,Momentum=0.3,Repetitions=1,ConvergenceSteps=10,BatchSize=50,"
"TestRepetitions=5,WeightDecay=0.01,Regularization=NONE");
TString trainingStrategyString("TrainingStrategy=");
trainingStrategyString += training0 + "|" + training1 + "|" + training2;
// TString trainingStrategyString
// ("TrainingStrategy=LearningRate=1e-1,Momentum=0.3,Repetitions=3,ConvergenceSteps=20,BatchSize=30,TestRepetitions=7,WeightDecay=0.0,L1=false,DropFraction=0.0,DropRepetitions=5");
TString nnOptions(
"!H:V:ErrorStrategy=SUMOFSQUARES:VarTransform=G:WeightInitialization=XAVIERUNIFORM:Architecture=CPU");
// TString nnOptions ("!H:V:VarTransform=Normalize:ErrorStrategy=CHECKGRADIENTS");
nnOptions.Append(":");
nnOptions.Append(layoutString);
nnOptions.Append(":");
nnOptions.Append(trainingStrategyString);
factory->BookMethod(dataloader, TMVA::Types::kDNN, "DNN_CPU", nnOptions); // NN
}
// Support Vector Machine
if (Use["SVM"])
factory->BookMethod( dataloader, TMVA::Types::kSVM, "SVM", "Gamma=0.25:Tol=0.001:VarTransform=Norm" );
// Boosted Decision Trees
if (Use["BDT"])
factory->BookMethod( dataloader, TMVA::Types::kBDT, "BDT",
"!H:!V:NTrees=100:MinNodeSize=1.0%:BoostType=AdaBoostR2:SeparationType=RegressionVariance:nCuts=20:PruneMethod=CostComplexity:PruneStrength=30" );
if (Use["BDTG"])
factory->BookMethod( dataloader, TMVA::Types::kBDT, "BDTG",
"!H:!V:NTrees=2000::BoostType=Grad:Shrinkage=0.1:UseBaggedBoost:BaggedSampleFraction=0.5:nCuts=20:MaxDepth=3:MaxDepth=4" );
// --------------------------------------------------------------------------------------------------
// Now you can tell the factory to train, test, and evaluate the MVAs
// Train MVAs using the set of training events
factory->TrainAllMethods();
// Evaluate all MVAs using the set of test events
factory->TestAllMethods();
// Evaluate and compare performance of all configured MVAs
factory->EvaluateAllMethods();
// --------------------------------------------------------------
// Save the output
outputFile->Close();
std::cout << "==> Wrote root file: " << outputFile->GetName() << std::endl;
std::cout << "==> TMVARegression is done!" << std::endl;
delete factory;
delete dataloader;
// Launch the GUI for the root macros
if (!gROOT->IsBatch()) TMVA::TMVARegGui( outfileName );
}
int main( int argc, char** argv )
{
// Select methods (don't look at this code - not of interest)
TString methodList;
for (int i=1; i<argc; i++) {
TString regMethod(argv[i]);
if(regMethod=="-b" || regMethod=="--batch") continue;
if (!methodList.IsNull()) methodList += TString(",");
methodList += regMethod;
}
TMVARegression(methodList);
return 0;
}
Author
Andreas Hoecker

Definition in file TMVARegression.C.