| 1 |
kukartse |
1.1 |
/**********************************************************************************
|
| 2 |
|
|
* Project : TMVA - a Root-integrated toolkit for multivariate data analysis *
|
| 3 |
|
|
* Package : TMVA *
|
| 4 |
|
|
* Exectuable: ClassApplication *
|
| 5 |
|
|
* *
|
| 6 |
|
|
* Test suit for comparison of Reader and standalone class outputs *
|
| 7 |
|
|
**********************************************************************************/
|
| 8 |
|
|
|
| 9 |
|
|
#include <vector>
|
| 10 |
|
|
|
| 11 |
|
|
void ClassApplication( TString myMethodList = "" )
|
| 12 |
|
|
{
|
| 13 |
|
|
cout << endl;
|
| 14 |
|
|
cout << "==> start ClassApplication" << endl;
|
| 15 |
|
|
|
| 16 |
|
|
const int Nmvas = 15;
|
| 17 |
|
|
const char* bulkname[Nmvas] = { "Likelihood", "LikelihoodD", "LikelihoodPCA", "LikelihoodMIX",
|
| 18 |
|
|
"HMatrix", "Fisher",
|
| 19 |
|
|
"MLP",
|
| 20 |
|
|
"BDT", "RuleFit",
|
| 21 |
|
|
"SVM_Gauss", "SVM_Poly", "SVM_Lin",
|
| 22 |
|
|
"FDA_MT", "FDA_MC", "FDA_GA" };
|
| 23 |
|
|
|
| 24 |
|
|
bool iuse[Nmvas] = { Nmvas*kFALSE };
|
| 25 |
|
|
|
| 26 |
|
|
// interpret input list
|
| 27 |
|
|
if (myMethodList != "") {
|
| 28 |
|
|
TList* mlist = TMVA::Tools::ParseFormatLine( myMethodList, " :," );
|
| 29 |
|
|
for (int imva=0; imva<Nmvas; imva++) if (mlist->FindObject( bulkname[imva] )) iuse[imva] = kTRUE;
|
| 30 |
|
|
delete mlist;
|
| 31 |
|
|
}
|
| 32 |
|
|
|
| 33 |
|
|
// create a set of variables and declare them to the reader
|
| 34 |
|
|
// - the variable names must corresponds in name and type to
|
| 35 |
|
|
// those given in the weight file(s) that you use
|
| 36 |
|
|
std::vector<std::string> inputVars;
|
| 37 |
|
|
inputVars.push_back( "var1+var2" );
|
| 38 |
|
|
inputVars.push_back( "var1-var2" );
|
| 39 |
|
|
inputVars.push_back( "var3" );
|
| 40 |
|
|
inputVars.push_back( "var4" );
|
| 41 |
|
|
|
| 42 |
|
|
// preload standalone class(es)
|
| 43 |
|
|
string dir = "weights/";
|
| 44 |
|
|
string prefix = "TMVAnalysis";
|
| 45 |
|
|
|
| 46 |
|
|
for (int imva=0; imva<Nmvas; imva++) {
|
| 47 |
|
|
if (iuse[imva]) {
|
| 48 |
|
|
TString cfile = dir + prefix + "_" + bulkname[imva] + ".class.C++";
|
| 49 |
|
|
|
| 50 |
|
|
cout << "=== Macro : Loading class file: " << cfile << endl;
|
| 51 |
|
|
|
| 52 |
|
|
// load the classifier's standalone class
|
| 53 |
|
|
gROOT->LoadMacro( cfile );
|
| 54 |
|
|
}
|
| 55 |
|
|
}
|
| 56 |
|
|
cout << "=== Macro : Classifier class loading successfully terminated" << endl;
|
| 57 |
|
|
|
| 58 |
|
|
// define classes
|
| 59 |
|
|
IClassifierReader* classReader[Nmvas] = { Nmvas*0 };
|
| 60 |
|
|
|
| 61 |
|
|
// ... and create them (and some histograms for the output)
|
| 62 |
|
|
int nbin = 100;
|
| 63 |
|
|
TH1* hist[Nmvas];
|
| 64 |
|
|
|
| 65 |
|
|
for (int imva=0; imva<Nmvas; imva++) {
|
| 66 |
|
|
if (iuse[imva]) {
|
| 67 |
|
|
cout << "=== Macro : Testing " << bulkname[imva] << endl;
|
| 68 |
|
|
if (bulkname[imva] == "Likelihood" ) {
|
| 69 |
|
|
classReader[imva] = new ReadLikelihood ( inputVars );
|
| 70 |
|
|
hist[imva] = new TH1F( "MVA_Likelihood", "MVA_Likelihood", nbin, 0, 1 );
|
| 71 |
|
|
}
|
| 72 |
|
|
if (bulkname[imva] == "LikelihoodD" ) {
|
| 73 |
|
|
classReader[imva] = new ReadLikelihoodD ( inputVars );
|
| 74 |
|
|
hist[imva] = new TH1F( "MVA_LikelihoodD", "MVA_LikelihoodD", nbin, 0, 1 );
|
| 75 |
|
|
}
|
| 76 |
|
|
if (bulkname[imva] == "LikelihoodPCA") {
|
| 77 |
|
|
classReader[imva] = new ReadLikelihoodPCA( inputVars );
|
| 78 |
|
|
hist[imva] = new TH1F( "MVA_LikelihoodPCA", "MVA_LikelihoodPCA", nbin, 0, 1 );
|
| 79 |
|
|
}
|
| 80 |
|
|
if (bulkname[imva] == "LikelihoodMIX") {
|
| 81 |
|
|
classReader[imva] = new ReadLikelihoodMIX( inputVars );
|
| 82 |
|
|
hist[imva] = new TH1F( "MVA_LikelihoodMIX", "MVA_LikelihoodMIX", nbin, 0, 1 );
|
| 83 |
|
|
}
|
| 84 |
|
|
if (bulkname[imva] == "HMatrix" ) {
|
| 85 |
|
|
classReader[imva] = new ReadHMatrix ( inputVars );
|
| 86 |
|
|
hist[imva] = new TH1F( "MVA_HMatrix", "MVA_HMatrix", nbin, -0.95, 1.55 );
|
| 87 |
|
|
}
|
| 88 |
|
|
if (bulkname[imva] == "Fisher" ) {
|
| 89 |
|
|
classReader[imva] = new ReadFisher ( inputVars );
|
| 90 |
|
|
hist[imva] = new TH1F( "MVA_Fisher", "MVA_Fisher", nbin, -4, 4 );
|
| 91 |
|
|
}
|
| 92 |
|
|
if (bulkname[imva] == "MLP" ) {
|
| 93 |
|
|
classReader[imva] = new ReadMLP ( inputVars );
|
| 94 |
|
|
hist[imva] = new TH1F( "MVA_MLP", "MVA_MLP", nbin, -0.25, 1.5 );
|
| 95 |
|
|
}
|
| 96 |
|
|
if (bulkname[imva] == "BDT" ) {
|
| 97 |
|
|
classReader[imva] = new ReadBDT ( inputVars );
|
| 98 |
|
|
hist[imva] = new TH1F( "MVA_BDT", "MVA_BDT", nbin, -0.8, 0.8 );
|
| 99 |
|
|
}
|
| 100 |
|
|
if (bulkname[imva] == "RuleFit" ) {
|
| 101 |
|
|
classReader[imva] = new ReadRuleFit ( inputVars );
|
| 102 |
|
|
hist[imva] = new TH1F( "MVA_RuleFitTMVA", "MVA_RuleFitTMVA", nbin, -2.0, 2.0 );
|
| 103 |
|
|
}
|
| 104 |
|
|
if (bulkname[imva] == "SVM_Gauss" ) {
|
| 105 |
|
|
classReader[imva] = new ReadSVM_Gauss ( inputVars );
|
| 106 |
|
|
hist[imva] = new TH1F( "MVA_SVM_Gauss", "MVA_SVM_Gauss", nbin, 0.0, 1.0 );
|
| 107 |
|
|
}
|
| 108 |
|
|
if (bulkname[imva] == "SVM_Lin" ) {
|
| 109 |
|
|
classReader[imva] = new ReadSVM_Lin ( inputVars );
|
| 110 |
|
|
hist[imva] = new TH1F( "MVA_SVM_Lin", "MVA_SVM_Lin", nbin, 0.0, 1.0 );
|
| 111 |
|
|
}
|
| 112 |
|
|
if (bulkname[imva] == "SVM_Poly" ) {
|
| 113 |
|
|
classReader[imva] = new ReadSVM_Poly ( inputVars );
|
| 114 |
|
|
hist[imva] = new TH1F( "MVA_SVM_Poly", "MVA_SVM_Poly", nbin, 0.0, 1.0 );
|
| 115 |
|
|
}
|
| 116 |
|
|
if (bulkname[imva] == "FDA_MT" ) {
|
| 117 |
|
|
classReader[imva] = new ReadFDA_MT ( inputVars );
|
| 118 |
|
|
hist[imva] = new TH1F( "MVA_FDA_MT", "MVA_FDA_MT", nbin, -2.0, 3.0 );
|
| 119 |
|
|
}
|
| 120 |
|
|
if (bulkname[imva] == "FDA_MC" ) {
|
| 121 |
|
|
classReader[imva] = new ReadFDA_MC ( inputVars );
|
| 122 |
|
|
hist[imva] = new TH1F( "MVA_FDA_MC", "MVA_FDA_MC", nbin, -2.0, 3.0 );
|
| 123 |
|
|
}
|
| 124 |
|
|
if (bulkname[imva] == "FDA_GA" ) {
|
| 125 |
|
|
classReader[imva] = new ReadFDA_GA ( inputVars );
|
| 126 |
|
|
hist[imva] = new TH1F( "MVA_FDA_GA", "MVA_FDA_GA", nbin, -2.0, 3.0 );
|
| 127 |
|
|
}
|
| 128 |
|
|
}
|
| 129 |
|
|
}
|
| 130 |
|
|
cout << "=== Macro : Class creation was successful" << endl;
|
| 131 |
|
|
|
| 132 |
|
|
// Prepare input tree (this must be replaced by your data source)
|
| 133 |
|
|
// in this example, there is a toy tree with signal and one with background events
|
| 134 |
|
|
// we'll later on use only the "signal" events for the test in this example.
|
| 135 |
|
|
//
|
| 136 |
|
|
TFile *input(0);
|
| 137 |
|
|
if (!gSystem->AccessPathName("./tmva_example.root")) {
|
| 138 |
|
|
// first we try to find tmva_example.root in the local directory
|
| 139 |
|
|
cout << "=== Macro : Accessing ./tmva_example.root" << endl;
|
| 140 |
|
|
input = TFile::Open("tmva_example.root");
|
| 141 |
|
|
}
|
| 142 |
|
|
else {
|
| 143 |
|
|
// second we try accessing the file via the web from
|
| 144 |
|
|
// http://root.cern.ch/files/tmva_example.root
|
| 145 |
|
|
cout << "=== Macro : Accessing tmva_example.root file from http://root.cern.ch/files" << endl;
|
| 146 |
|
|
cout << "=== Macro : For faster startup you may consider downloading it into you local directory" << endl;
|
| 147 |
|
|
input = TFile::Open("http://root.cern.ch/files/tmva_example.root");
|
| 148 |
|
|
}
|
| 149 |
|
|
|
| 150 |
|
|
if (!input) {
|
| 151 |
|
|
cout << "ERROR: could not open data file" << endl;
|
| 152 |
|
|
exit(1);
|
| 153 |
|
|
}
|
| 154 |
|
|
|
| 155 |
|
|
//
|
| 156 |
|
|
// prepare the tree
|
| 157 |
|
|
// - here the variable names have to corresponds to your tree
|
| 158 |
|
|
// - you can use the same variables as above which is slightly faster,
|
| 159 |
|
|
// but of course you can use different ones and copy the values inside the event loop
|
| 160 |
|
|
//
|
| 161 |
|
|
TTree* theTree = (TTree*)input->Get("TreeS");
|
| 162 |
|
|
cout << "=== Macro : Loop over signal sample" << endl;
|
| 163 |
|
|
|
| 164 |
|
|
// the references to the variables
|
| 165 |
|
|
float var1, var2, var3, var4;
|
| 166 |
|
|
float userVar1, userVar2;
|
| 167 |
|
|
theTree->SetBranchAddress( "var1", &userVar1 );
|
| 168 |
|
|
theTree->SetBranchAddress( "var2", &userVar2 );
|
| 169 |
|
|
theTree->SetBranchAddress( "var3", &var3 );
|
| 170 |
|
|
theTree->SetBranchAddress( "var4", &var4 );
|
| 171 |
|
|
|
| 172 |
|
|
cout << "=== Macro : Processing total of " << theTree->GetEntries() << " events ... " << endl;
|
| 173 |
|
|
|
| 174 |
|
|
std::vector<double>* inputVec = new std::vector<double>( 4 );
|
| 175 |
|
|
for (Long64_t ievt=0; ievt<theTree->GetEntries();ievt++) {
|
| 176 |
|
|
|
| 177 |
|
|
if (ievt%1000 == 0) cout << "=== Macro : ... processing event: " << ievt << endl;
|
| 178 |
|
|
|
| 179 |
|
|
theTree->GetEntry(ievt);
|
| 180 |
|
|
|
| 181 |
|
|
var1 = userVar1 + userVar2;
|
| 182 |
|
|
var2 = userVar1 - userVar2;
|
| 183 |
|
|
|
| 184 |
|
|
(*inputVec)[0] = var1;
|
| 185 |
|
|
(*inputVec)[1] = var2;
|
| 186 |
|
|
(*inputVec)[2] = var3;
|
| 187 |
|
|
(*inputVec)[3] = var4;
|
| 188 |
|
|
|
| 189 |
|
|
// loop over all booked classifiers
|
| 190 |
|
|
for (int imva=0; imva<Nmvas; imva++) {
|
| 191 |
|
|
|
| 192 |
|
|
if (iuse[imva]) {
|
| 193 |
|
|
|
| 194 |
|
|
// retrive the classifier responses
|
| 195 |
|
|
double retval = classReader[imva]->GetMvaValue( *inputVec );
|
| 196 |
|
|
hist[imva]->Fill( retval, 1.0 );
|
| 197 |
|
|
}
|
| 198 |
|
|
}
|
| 199 |
|
|
}
|
| 200 |
|
|
|
| 201 |
|
|
cout << "=== Macro : Event loop done! " << endl;
|
| 202 |
|
|
|
| 203 |
|
|
TFile *target = new TFile( "ClassApp.root","RECREATE" );
|
| 204 |
|
|
for (int imva=0; imva<Nmvas; imva++) {
|
| 205 |
|
|
if (iuse[imva]) {
|
| 206 |
|
|
hist[imva]->Write();
|
| 207 |
|
|
}
|
| 208 |
|
|
}
|
| 209 |
|
|
cout << "=== Macro : Created target file: " << target->GetName() << endl;
|
| 210 |
|
|
target->Close();
|
| 211 |
|
|
|
| 212 |
|
|
delete target;
|
| 213 |
|
|
delete inputVec;
|
| 214 |
|
|
|
| 215 |
|
|
cout << "==> ClassApplication is done!" << endl << endl;
|
| 216 |
|
|
}
|
