new HOWTO macros

diff --git a/STAT/Macros/testInterpolator.C b/STAT/Macros/testInterpolator.C
index 1bad41ef24847a24b65a9cb0acb53a4686ee6f54..8de2cd75e08621bf98b2721c06ca3ec5efd67dbf 100644 (file)
--- a/STAT/Macros/testInterpolator.C
+++ b/STAT/Macros/testInterpolator.C
@@ -1,32 +1,154 @@
-void testInterpolator()
+const Int_t ndim = 2;
+Double_t testInterpolator(const Int_t nstat = 100000)
 {
-       gStyle->SetOptStat(0);
+// Macro for testing the TKDInterpolator.
+// 
+// The function which it is interpolated is an uncorrelated landau
+// distribution in "ndim" dimensions. The function returns the chi2 of
+// the interpolation.
+// 
+// Parameters
+//     nstat - number of points to be used for training
+//     kBuild - on/off generate data
+//     kTransform - on/off outliers compresion
+//     kPCA - on/off pricipal component analysis 
+
+       const Bool_t kBuild = 1;
+       const Bool_t kTransform = 1;
+       const Bool_t kPCA = 1;
        gStyle->SetPalette(1);
        
-       Int_t npoints = 301;
-       Int_t bsize = 10;
+       Double_t pntTrain[ndim], pntTest[ndim], pntRotate[ndim];
+       Double_t pdf;
+       TFile *f = 0x0, *fEval = 0x0;
+       TTree *t = 0x0, *tEval = 0x0;
+
+
+       // build data
+       if(kBuild){
+               printf("build data ... \n");
+               f = TFile::Open(Form("%dD_LL.root", ndim), "RECREATE");
+               t = new TTree("db", "Log-Log database");
+               for(int idim=0; idim<ndim; idim++) t->Branch(Form("x%d", idim), &pntTrain[idim], Form("x%d/D", idim));
+               
+               fEval = TFile::Open(Form("%dD_Eval.root", ndim), "RECREATE");
+               tEval = new TTree("db", "Evaluation database");
+               for(int idim=0; idim<ndim; idim++) tEval->Branch(Form("x%d", idim), &pntTest[idim], Form("x%d/D", idim));
+               
+               for(int istat=0; istat<nstat; istat++){
+                       for(int idim=0; idim<ndim; idim++) pntTrain[idim] = gRandom->Landau(5.);
+                       if(!(istat%3)){ // one third of the statistics is for testing
+                               memcpy(pntTest, pntTrain, ndim*sizeof(Double_t));
+                               tEval->Fill();
+                               continue;
+                       }
+                       if(kTransform)
+                               for(int idim=0; idim<ndim; idim++)
+                                       if(pntTrain[idim] > 0.) pntTrain[idim] = TMath::Log(pntTrain[idim]);
+                                       else pntTrain[idim] = 0.;
+                       t->Fill();
+               }
+               f->cd();
+               t->Write();
+               f->Flush();
+               
+               fEval->cd();
+               tEval->Write();
+               fEval->Flush();
+       } else {// link data
+               printf("link data ... \n");
+               f = TFile::Open(Form("%dD_LL.root", ndim));
+               t = (TTree*)f->Get("db");
+               for(int idim=0; idim<ndim; idim++) t->SetBranchAddress(Form("x%d", idim), &pntTrain[idim]);
+               
+               fEval = TFile::Open(Form("%dD_Eval.root", ndim));
+               tEval = (TTree*)fEval->Get("db");
+               for(int idim=0; idim<ndim; idim++) tEval->SetBranchAddress(Form("x%d", idim), &pntTest[idim]);
+       }
+
        
-       Float_t *data0 =  new Float_t[npoints*2];
-       Float_t *data[2];
-       data[0] = &data0[0];
-       data[1] = &data0[npoints];
-       for (Int_t i=0;i<npoints;i++) {
-               data[1][i]= gRandom->Gaus(.5, .1);
-               data[0][i]= gRandom->Gaus(.5, .1);
+       // do principal component analysis (PCA)
+       TPrincipal princ(ndim, "N");
+       if(kPCA && kBuild){
+               printf("do principal component analysis (PCA) ... \n");
+               f->cd();
+               TTree *tt = new TTree("db1", "PCA database");
+               for(int idim=0; idim<ndim; idim++) tt->Branch(Form("x%d", idim), &pntRotate[idim], Form("x%d/D", idim));
+               for(int ientry=0; ientry<t->GetEntries(); ientry++){
+                       t->GetEntry(ientry);
+                       princ.AddRow(pntTrain);
+               }
+               princ.MakePrincipals();
+               for(int ientry=0; ientry<t->GetEntries(); ientry++){
+                       t->GetEntry(ientry);
+                       princ.X2P(pntTrain, pntRotate);
+                       tt->Fill();
+               }
+               tt->Write();
+               f->Flush();
+               for(int idim=0; idim<ndim; idim++) tt->SetBranchAddress(Form("x%d", idim), &pntTrain[idim]);
+               t = tt;
        }
-       TKDInterpolator *in = new TKDInterpolator(npoints, 2, bsize, data);
+       gROOT->cd();
+       
+       // do interpolation
+       printf("do interpolation ... \n");
+       Double_t pdf, pdf_estimate, chi2;
+       TString vl = "x0";
+       for(int idim=1; idim<ndim; idim++) vl+=Form(":x%d", idim);
+       TKDInterpolator in(t, vl.Data(), "", 200.);
+       chi2 = 0.;
+/*     for(int ip=0; ip<tEval->GetEntries(); ip++){
+               tEval->GetEntry(ip);
+               printf("\nEval %d\n", ip);*/
+       TH1 *h1 = new TH2F("h1", "", 50, 0., 100., 50, 0., 100.);
+       TH1 *h2 = new TH2F("h2", "", 50, 0., 100., 50, 0., 100.);
+       TAxis *ax = h2->GetXaxis(), *ay = h2->GetYaxis();
+       for(int ix=2; ix<ax->GetNbins(); ix++){
+               pntTest[0] = ax->GetBinCenter(ix);
+       for(int iy=2; iy<ay->GetNbins(); iy++){
+               pntTest[1] = ay->GetBinCenter(iy);
+               memcpy(pntTrain, pntTest, ndim*sizeof(Double_t));
 
-       TH2 *hS = new TH2F("hS", "", 10, .3, .7, 10, .3, .7);
-       TAxis *ax = hS->GetXaxis(), *ay = hS->GetYaxis();
-       Float_t p[2], eval;
-       for(int ix=1; ix<=ax->GetNbins(); ix++){
-               p[0] = ax->GetBinCenter(ix);
-               for(int iy=1; iy<=ay->GetNbins(); iy++){
-                       p[1] = ay->GetBinCenter(iy);
-                       eval = in->Eval(p);
-                       //printf("x %f y %f eval %f [%d]\n", p[0], p[1], eval, TMath::IsNaN(eval));
-                       if(!TMath::IsNaN(eval)) hS->SetBinContent(ix, iy, eval);
+               if(kTransform)
+                       for(int idim=0; idim<ndim; idim++)
+                               if(pntTrain[idim] > 0.) pntTrain[idim] = TMath::Log(pntTrain[idim]);
+                               else pntTrain[idim] = 0.;
+               
+               if(kPCA){
+                       princ.X2P(pntTrain, pntRotate);
+                       memcpy(pntTrain, pntRotate, ndim*sizeof(Double_t));
                }
+
+               pdf_estimate = in.Eval(pntTrain, 30);
+               // calculate chi2
+               if(kTransform)
+                       for(int idim=0; idim<ndim; idim++)
+                               if(pntTest[idim] > 0.) pdf_estimate /= pntTest[idim];
+                               else continue; 
+               
+               h1->SetBinContent(ix, iy, pdf_estimate);
+               
+               pdf = 1.; for(int idim=0; idim<ndim; idim++) pdf *= TMath::Landau(pntTest[idim], 5.);
+               h2->SetBinContent(ix, iy, pdf);
+               pdf_estimate -= pdf;
+               chi2 += pdf_estimate*pdf_estimate/pdf;
+       }}
+       f->Close(); delete f;
+       fEval->Close(); delete fEval;
+       
+       // results presentation
+       printf("chi2 = %f\n", chi2);
+       TCanvas *c = 0x0;
+       if(!(c = (TCanvas*)gROOT->FindObject("c"))){
+               c = new TCanvas("c", "", 10, 10, 900, 500);
+               c->Divide(2, 1);
        }
-       hS->Draw("lego2");
-}
\ No newline at end of file
+       c->cd(1);
+       h1->Draw("lego2"); h1->GetZaxis()->SetRangeUser(1.e-9, 5.e-2); gPad->SetLogz(); gPad->Modified(); gPad->Update();
+       
+       c->cd(2);
+       h2->Draw("lego2"); h2->GetZaxis()->SetRangeUser(1.e-9, 5.e-2); gPad->SetLogz(); gPad->Modified(); gPad->Update();
+       return chi2;
+}
+

diff --git a/STAT/Macros/testKNN.C b/STAT/Macros/testKNN.C
new file mode 100644 (file)
index 0000000..e526bfd
--- /dev/null
+++ b/STAT/Macros/testKNN.C
@@ -0,0 +1,50 @@
+const Float_t p[]={1.4, -.6}; //}{1.7, -.4};
+void testKNN(const Float_t *p, const int kNN=20)
+{
+// Draw "kNN" nearest neighbours of point "p". The distance (in the L1
+// metric) is encoded in the color code.
+// To build the data refere to function build().
+
+       TFile::Open("2D_Gauss.root");
+       TKDInterpolator in(db, "x0:x1", "x0>-1.5&&x0<2.&&x1>-2.&&x1<2.", 300);
+       in.DrawNode(in.FindNode(p)-in.GetNNodes());
+       
+       TMarker *mp = new TMarker(p[0], p[1], 3);
+       mp->SetMarkerColor(4);
+       mp->Draw();
+               
+       Int_t *index, color;
+       Float_t d, d0, pknn[2];
+       in.FindNearestNeighbors(p, kNN, index, d0);
+       TMarker *ma = new TMarker[kNN];
+       for(int ip=0; ip<kNN; ip++){
+               in.GetDataPoint(index[ip], pknn);
+               d = TMath::Abs(p[0]-pknn[0]) + TMath::Abs(p[1]-pknn[1]);
+               ma[ip].SetMarkerStyle(4);
+               color = 101 - Int_t(50. * d/d0);
+               ma[ip].SetMarkerColor(color);
+               ma[ip].DrawMarker(pknn[0], pknn[1]);
+       }
+}
+
+void build(const Int_t ndim = 2, const Int_t nstat = 100000)
+{
+// Build "nstat" data points in "ndim" dimensions taken from an
+// uncorrelated 2D Gauss distribution.
+       
+
+       printf("build data ... \n");
+       Double_t pntTrain[ndim];
+       f = TFile::Open(Form("%dD_Gauss.root", ndim), "RECREATE");
+       t = new TTree("db", "gauss database");
+       for(int idim=0; idim<ndim; idim++) t->Branch(Form("x%d", idim), &pntTrain[idim], Form("x%d/D", idim));
+       
+       for(int istat=0; istat<nstat; istat++){
+               for(int idim=0; idim<ndim; idim++) pntTrain[idim] = gRandom->Gaus();
+               t->Fill();
+       }
+       f->cd();
+       t->Write();
+       f->Close();
+       delete f;
+}
\ No newline at end of file