int main(){
std::vector<MCSample> Samples;
PkPhi.FileName="../PkphiSingle.root";
PkPhi.Description="#Lambda_b -> p K #phi background";
Samples.push_back(PkPhi);
Rare.FileName="../RareMCSingle.root";
Rare.Description="#eta' -> #pi #pi #gamma";
Samples.push_back(Rare);
Rare2.FileName="../Rare2MCSingle.root";
Rare2.Description="#eta' -> #pi #pi #eta";
Samples.push_back(Rare2);
Control.FileName="../CCMCSingle.root";
Control.Description="Control Channel";
Samples.push_back(Control);
for(auto &Sample : Samples){
TreeReader * Reader = new TreeReader("DecayTree");
Reader->AddFile(Sample.FileName.data());
Reader->Initialize();
Long64_t N=Reader->GetEntries();
Sample.CC=new CorrelationCoeff(Sample.Description);
std::string HeadMassName="Lambda_b0_DTF_MF";
Reader->GetEntry(0);
try{
Reader->GetValue("Lambda_b0_DTF_MF");
}catch(...){
HeadMassName="Bu_DTF_MF";
}
for(int i=0;i<N;++i){
Reader->GetEntry(i);
double HeadMass=Reader->GetValue(HeadMassName.data());
double EtaMass=Reader->GetValue("eta_prime_MM");
if(HeadMassName.compare("Lambda_b0_DTF_MF")==0){
if(HeadMass>5200.0&&HeadMass<6000.0&&EtaMass>880.0&&EtaMass<1040.0){
Sample.CC->Fill(HeadMass,EtaMass);
}
}else{
if(HeadMass>5000.0&&HeadMass<5500.0&&EtaMass>880.0&&EtaMass<1040.0){
Sample.CC->Fill(HeadMass,EtaMass);
}
}
}
double NBootStraps=N*10;
std::cout<<"After "<<NBootStraps<<" Boostraps"<<std::endl;
std::cout<<Sample.Description<<" Coefficient = "<<Sample.CC->GetCC()<<" +/- "<<Sample.CC->GetBootstrapError(NBootStraps)<<std::endl;
}
TFile * Output = new TFile("Output.root","RECREATE");
for(auto &Sample : Samples){
Sample.CC->GetBootStrapHist(0)->Write();
Sample.CC->GetCorrelationPlot()->Write();
}
}
int main(int argc, char** argv)
{
string dataType = "12";
//if(argc > 1) if((string)argv[1] == "11") dataType = "11";
NeuroBayesTeacher* nb = NeuroBayesTeacher::Instance();
nb->NB_DEF_TASK("CLASSIFICATION");
//setup network topology
int nvar = 20; // Set this to number of inputs to your NN
char ** varnames = new char*[nvar];
varnames[0] = "chi2_DTF";
varnames[1] = "Lb_TAU";
varnames[2] = "Lb_DIRA_OWNPV";
varnames[3] = "Lb_IPCHI2_OWNPV";
varnames[4] = "max_mu_IPCHI2_OWNPV";
varnames[5] = "min_mu_TRACKCHI2";
varnames[6] = "min_mu_PID";
varnames[7] = "min_mu_PID";
varnames[8] = "LL_Lambda0_IPCHI2_OWNPV";
varnames[9] = "LL_Lambda0_FDCHI2_OWNPV";
varnames[10] = "LL_Lambda0_PT";
varnames[11] = "DD_Lambda0_IPCHI2_OWNPV";
varnames[12] = "DD_Lambda0_FDCHI2_OWNPV";
varnames[13] = "DD_Lambda0_PT";
varnames[14] = "DD_pplus_IPCHI2_OWNPV";
varnames[15] = "DD_piminus_IPCHI2_OWNPV";
varnames[16] = "DD_piminus_PT";
varnames[17] = "LL_pplus_IPCHI2_OWNPV";
varnames[18] = "LL_piminus_IPCHI2_OWNPV";
varnames[19] = "LL_piminus_PT";
nb->NB_DEF_NODE1(nvar+1);
nb->NB_DEF_NODE2(nvar); // nodes in hidden layer
nb->NB_DEF_NODE3(1); // nodes in output layer
nb->NB_DEF_TASK("CLA"); // binominal classification
nb->NB_DEF_PRE(822);
// nb->NB_DEF_PRE(812);
nb->NB_DEF_REG("REG"); // 'OFF','REG' (def) ,'ARD','ASR','ALL'
nb->NB_DEF_LOSS("ENTROPY"); // 'ENTROPY'(def),'QUADRATIC'
nb->NB_DEF_METHOD("BFGS");
nb->NB_DEF_SHAPE("DIAG");
nb->NB_DEF_LEARNDIAG(1);
nb->NB_DEF_RTRAIN(1.0); // use 70% of events for training
// nb->NB_DEF_EPOCH(200); // weight update after n events
nb->NB_DEF_SPEED(2.0); // multiplicative factor to enhance global learning speed
nb->NB_DEF_MAXLEARN(1.0); // multiplicative factor to limit the global learning speed in any direction, this number should be smaller than NB_DEF_SPEED
nb->NB_DEF_ITER(100); // number of training iteration
//nb->NB_DEF_ITER(0); // number of training iteration
//int i = 4701;
//int j = 29;
//nb->NB_RANVIN(i,j,2); // random number seed initialisation, i has to be an odd number, the third argument is a debugging flag
nb->SetOutputFile(("expert_"+dataType+".nb").c_str()); // expert file
SetupNNPrepro(nb);
// MC
TreeReader* reader = new TreeReader("tree");
reader->AddFile("/afs/cern.ch/work/p/pluca/Lmumu/weighted/Lb2Lmumu_MC_Pythia8_NBweighted_new.root");
reader->Initialize();
// We take all signal and 20% of background
nb->SetTarget(1);
int ntot = reader->GetEntries();
int npassedMC = 0;
cout << "Read in " << ntot << " events" << endl;
int nstepMC = 5;
//if(dataType=="11") nstepMC = 5;
TFile ofile("/afs/cern.ch/work/p/pluca/Lmumu/weighted/samplesMVA_"+(TString)dataType+".root","recreate");
TTree * sigTrainSample = new TTree("sigTrainSample","");
reader->BranchNewTree(sigTrainSample);
TTree * sigTestSample = new TTree("sigTestSample","");
reader->BranchNewTree(sigTestSample);
for(int event = 0; event < ntot; event++)
{
reader->GetEntry(event);
if( TrueID(reader) && TriggerPassed(reader))
{
if( event%nstepMC==0 && npassedMC <= 4e4 )
{
npassedMC++;
float InputArray[nvar+1];
fillInputArray(reader,InputArray);
if(isnan(InputArray[0])) continue;
nb->SetWeight(reader->GetValue("Lb_weight"));
nb->SetNextInput(nvar,InputArray);
sigTrainSample->Fill();
}
else sigTestSample->Fill();
}
}
// Data
TreeReader* reader2 = new TreeReader("tree");
reader2->AddFile("/afs/cern.ch/work/p/pluca/Lmumu/weighted/Lb2Lmumu_CL_NBweighted.root");
reader2->Initialize();
TTree * bkgTrainSample = new TTree("bkgTrainSample","");
reader2->BranchNewTree(bkgTrainSample);
TTree * bkgTestSample = new TTree("bkgTestSample","");
reader2->BranchNewTree(bkgTestSample);
nb->SetTarget(0);
int ntot2 = reader2->GetEntries();
int npassed = 0;
cout << "Read in " << ntot2 << " events" << endl;
int nstep = 2;
//if(dataType=="11") nstep = 2;
for(int event = 0; event < ntot2; event++)
{
reader2->GetEntry(event);
double massLb = reader2->GetValue("Lb_MassConsLambda_M",0);
double massJpsi = reader2->GetValue("J_psi_1S_MM");
if(massLb > 6000 && TMath::Abs(massJpsi - 3096) > 100 && TMath::Abs(massJpsi - 3686) > 90 && TriggerPassed(reader2))
{
if(event%nstep==0 && npassed <=4e4)
{
float InputArray[100];
npassed++;
fillInputArray(reader2,InputArray);
if(isnan(InputArray[0])) continue;
nb->SetWeight(1.);
nb->SetNextInput(nvar,InputArray);
bkgTrainSample->Fill();
}
else bkgTestSample->Fill();
}
}
bkgTrainSample->Write();
bkgTestSample->Write();
sigTestSample->Write();
sigTrainSample->Write();
ofile.Close();
cout << "\nData used = " << npassed << ", MC used = " << npassedMC << endl;
cout << "Train the Network\n" << endl;
nb->TrainNet();
nb->nb_correl_signi(varnames,"correl_signi.txt","correl_signi.html");
cout << "\nData used = " << npassed << ", MC used = " << npassedMC << endl;
return 0;
}
