Skip to content

New Methods #192

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
pouryavakilipour wants to merge 2 commits into from
Closed

New Methods #192

Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
e82d488
it is a homework
pouryavakilipour Jun 28, 2016
43f354b
// I've added two new mwthods
pouryavakilipour Jul 11, 2016
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Prev Previous commit
// I've added two new mwthods
  • Loading branch information
@pouryavakilipour
pouryavakilipour committed Jul 11, 2016
commit 43f354baad73f1cc2fe6424c1ab57230b8a2e90d
283 changes: 273 additions & 10 deletions tmva/tmva/src/Factory.cxx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -52,6 +52,8 @@
#include "TText.h"
#include "TLegend.h"
#include "TGraph.h"
#include "TMultiGraph.h"
#include "TSpline.h"
#include "TStyle.h"
#include "TMatrixF.h"
#include "TMatrixDSym.h"
Expand Down Expand Up @@ -98,11 +100,6 @@

#include <TCanvas.h>

Double_t auc[NumFolds];
Int_t y[NumFolds];
for(Int_t i = 0; i < NumFolds; i++){
y[i] = i+1;
}

const Int_t MinNoTrainingEvents = 10;
//const Int_t MinNoTestEvents = 1;
Expand Down Expand Up @@ -2084,6 +2081,7 @@ float TMVA::Factory::CrossValidate(DataLoader * loader, Types::EMVA theMethod, T
EvaluateAllMethods();

ROCs.push_back(GetROCIntegral(seedloader->GetName(), methodTitle));
// here is where i have to GET the ROCCurve

TMVA::MethodBase * smethod = dynamic_cast<TMVA::MethodBase*>(fMethodsMap[seedloader->GetName()][0][0]);
TMVA::ResultsClassification * sresults = (TMVA::ResultsClassification*)smethod->Data()->GetResults(smethod->GetMethodName(), Types::kTesting, Types::kClassification);
Expand Down Expand Up @@ -2117,15 +2115,280 @@ float TMVA::Factory::CrossValidate(DataLoader * loader, Types::EMVA theMethod, T
for(UInt_t l=0; l<ROCs.size(); ++l){
if (rocIntegrals) rocIntegrals[l] = ROCs.at(l);
std::cout << "Fold " << l+1 << " ROCIntegral: " << ROCs.at(l) << std::endl;
auc[l] = ROCs.at(l);
}
std::cout << "Average ROCIntegral: " << sumFOM/(double)NumFolds << std::endl;
}

TCanvas c1 = new TCanvas("c1","AUC Histo");
gr = new TGraph(NumFolds, auc, y);
gr->Draw("AC*");

return sumFOM/(double)NumFolds;


}


// Histogram of ROC Integrals

void TMVA::Factory::CrossValidateAUCHisto(DataLoader * loader, Types::EMVA theMethod, TString methodTitle, const char *theOption, bool optParams, int NumFolds, bool remakeDataSet)
{

Double_t AUC[NumFolds];
Double_t Y[NumFolds];
for(Int_t i = 0; i < NumFolds; i++){
Y[i] = i + 1.00;
}

//bool optParams = true;

if(remakeDataSet){
//loader->ValidationKFoldSet();
loader->MakeKFoldDataSet(NumFolds);
}

std::vector<float> parameterPerformance;

const int nbits = loader->DefaultDataSetInfo().GetNVariables();
std::vector<TString> varNames = loader->DefaultDataSetInfo().GetListOfVariables();

if(optParams){

std::vector<std::map<TString,Double_t> > foldParameters;

//loader->ValidationKFoldSet();

for(Int_t i=0; i<NumFolds; ++i){
Event::SetIsTraining(kTRUE);
TString optTitle = methodTitle;
optTitle += "_opt";
optTitle += i;

loader->PrepareTrainingAndTestTree(i, TMVA::Types::kTraining);

TMVA::DataLoader * seedloader = new TMVA::DataLoader(optTitle);

for(int index = 0; index<nbits; index++){
seedloader->AddVariable(varNames.at(index), 'F');
}

VIDataLoaderCopy(seedloader,loader);

MethodBase* mva = BookMethod(seedloader, theMethod, methodTitle, theOption);

foldParameters.push_back(mva->OptimizeTuningParameters("ROCIntegral","Minuit"));

this->DeleteAllMethods();

fMethodsMap.clear();
}

TString optionsString;

for(UInt_t t=0; t<foldParameters.size(); t++){
optionsString = theOption;
optionsString += ":";
std::map<TString,Double_t>::iterator it;
for(it=foldParameters.at(t).begin(); it!=foldParameters.at(t).end(); it++){
optionsString += it->first;
optionsString += "=";
optionsString += it->second;
if(it!=--foldParameters.at(t).end()){ optionsString += ":"; }
}
parameterPerformance.push_back(CrossValidate(loader, theMethod, methodTitle, optionsString, false, false));
}
}

std::vector<float> ROCs;

if(!optParams){

for(Int_t j=0; j<NumFolds; ++j){
TString foldTitle = methodTitle;
foldTitle += "_fold";
foldTitle += j+1;

loader->PrepareTrainingAndTestTree(j, TMVA::Types::kTesting);

TMVA::DataLoader * seedloader = new TMVA::DataLoader(foldTitle);

for(int index = 0; index<nbits; index++){
seedloader->AddVariable(varNames.at(index), 'F');
}

VIDataLoaderCopy(seedloader,loader);

BookMethod(seedloader, theMethod, methodTitle, theOption);

TrainAllMethods();
TestAllMethods();
EvaluateAllMethods();

ROCs.push_back(GetROCIntegral(seedloader->GetName(), methodTitle));

TMVA::MethodBase * smethod = dynamic_cast<TMVA::MethodBase*>(fMethodsMap[seedloader->GetName()][0][0]);
TMVA::ResultsClassification * sresults = (TMVA::ResultsClassification*)smethod->Data()->GetResults(smethod->GetMethodName(), Types::kTesting, Types::kClassification);
sresults->Clear();
sresults->Delete();
delete sresults;
fgTargetFile->cd();
fgTargetFile->Delete(seedloader->GetName());
fgTargetFile->Delete(Form("%s;1",seedloader->GetName()));
fgTargetFile->Flush();
gSystem->Exec(Form("rm -rf %s", seedloader->GetName()));

this->DeleteAllMethods();

fMethodsMap.clear();
}
}

float sumFOM = 0.0;

for(UInt_t k=0; k<ROCs.size(); ++k){
sumFOM += ROCs.at(k);
}

if(optParams){
for(UInt_t t=0; t<parameterPerformance.size(); ++t){
std::cout << "Parameters " << t+1 << " performance: " << parameterPerformance.at(t) << std::endl;
}
}
else{
for(UInt_t l=0; l<ROCs.size(); ++l){
AUC[l] = ROCs.at(l);
}
}

Double_t min = 0.0;
Double_t max = 0.0;
min = AUC[0];
max = AUC[0];

for(Int_t i = 1; i < NumFolds; i++){
if(AUC[i] > max){ max = AUC[i]; }
if(AUC[i] < min){ min = AUC[i]; }
}
TCanvas c ("c", "Histogram");
TH1D histo("AUC Distribution", "Histo", 100, min - 0.01, max + 0.01);
for(Int_t i = 0; i < NumFolds; i++){
histo.Fill(AUC[i], Y[i]);
}
histo.SetFillColor(90);
histo.SetTitle(" AUC Distribution ");
histo.Draw("HIST");
c.Print("AUCHisto.pdf");

}
// Different ROCs in a Graph

void TMVA::Factory::CrossValidatePlotROC(DataLoader * loader, Types::EMVA theMethod, TString methodTitle, const char *theOption, bool optParams, int NumFolds, bool remakeDataSet)
{

//bool optParams = true;

if(remakeDataSet){
//loader->ValidationKFoldSet();
loader->MakeKFoldDataSet(NumFolds);
}

std::vector<float> parameterPerformance;

const int nbits = loader->DefaultDataSetInfo().GetNVariables();
std::vector<TString> varNames = loader->DefaultDataSetInfo().GetListOfVariables();

if(optParams){

std::vector<std::map<TString,Double_t> > foldParameters;

//loader->ValidationKFoldSet();

for(Int_t i=0; i<NumFolds; ++i){
Event::SetIsTraining(kTRUE);
TString optTitle = methodTitle;
optTitle += "_opt";
optTitle += i;

loader->PrepareTrainingAndTestTree(i, TMVA::Types::kTraining);

TMVA::DataLoader * seedloader = new TMVA::DataLoader(optTitle);

for(int index = 0; index<nbits; index++){
seedloader->AddVariable(varNames.at(index), 'F');
}

VIDataLoaderCopy(seedloader,loader);

MethodBase* mva = BookMethod(seedloader, theMethod, methodTitle, theOption);

foldParameters.push_back(mva->OptimizeTuningParameters("ROCIntegral","Minuit"));

this->DeleteAllMethods();

fMethodsMap.clear();
}

TString optionsString;

for(UInt_t t=0; t<foldParameters.size(); t++){
optionsString = theOption;
optionsString += ":";
std::map<TString,Double_t>::iterator it;
for(it=foldParameters.at(t).begin(); it!=foldParameters.at(t).end(); it++){
optionsString += it->first;
optionsString += "=";
optionsString += it->second;
if(it!=--foldParameters.at(t).end()){ optionsString += ":"; }
}
parameterPerformance.push_back(CrossValidate(loader, theMethod, methodTitle, optionsString, false, false));
}
}

TCanvas *c = new TCanvas("c", "MultiGraph", 700, 500);
TMultiGraph *mg = new TMultiGraph();

if(!optParams){

for(Int_t j=0; j<NumFolds; ++j){
TString foldTitle = methodTitle;
foldTitle += "_fold";
foldTitle += j+1;

loader->PrepareTrainingAndTestTree(j, TMVA::Types::kTesting);

TMVA::DataLoader * seedloader = new TMVA::DataLoader(foldTitle);

for(int index = 0; index<nbits; index++){
seedloader->AddVariable(varNames.at(index), 'F');
}

VIDataLoaderCopy(seedloader,loader);

BookMethod(seedloader, theMethod, methodTitle, theOption);

TrainAllMethods();
TestAllMethods();
EvaluateAllMethods();

GetROCCurve(seedloader->GetName(), "BDT", true) -> SetLineColor(j*10);
mg -> Add(GetROCCurve(seedloader->GetName(), "BDT", true));

TMVA::MethodBase * smethod = dynamic_cast<TMVA::MethodBase*>(fMethodsMap[seedloader->GetName()][0][0]);
TMVA::ResultsClassification * sresults = (TMVA::ResultsClassification*)smethod->Data()->GetResults(smethod->GetMethodName(), Types::kTesting, Types::kClassification);
sresults->Clear();
sresults->Delete();
delete sresults;
fgTargetFile->cd();
fgTargetFile->Delete(seedloader->GetName());
fgTargetFile->Delete(Form("%s;1",seedloader->GetName()));
fgTargetFile->Flush();
gSystem->Exec(Form("rm -rf %s", seedloader->GetName()));

this->DeleteAllMethods();

fMethodsMap.clear();
}
}

mg -> Draw("al");
mg -> GetXaxis() -> SetTitle(" Signal Efficiency ");
mg -> GetYaxis() -> SetTitle(" Background Rejection ");
c -> Print("ROC.pdf");

}