JETAN/AliDAJetFinder.cxx

   1
   2 //  **************************************************************************
   3 //  * Copyright(c) 1998-1999, ALICE Experiment at CERN, All rights reserved. *
   4 //  *                                                                        *
   5 //  * Author: The ALICE Off-line Project.                                    *
   6 //  * Contributors are mentioned in the code where appropriate.              *
   7 //  *                                                                        *
   8 //  * Permission to use, copy, modify and distribute this software and its   *
   9 //  * documentation strictly for non-commercial purposes is hereby granted   *
  10 //  * without fee, provided that the above copyright notice appears in all   *
  11 //  * copies and that both the copyright notice and this permission notice   *
  12 //  * appear in the supporting documentation. The authors make no claims     *
  13 //  * about the suitability of this software for any purpose. It is          *
  14 //  * provided "as is" without express or implied warranty.                  *
  15 //  **************************************************************************
  16
  17 //-----------------------------------------------------------------------------------
  18 // Jet finder based on Deterministic Annealing
  19 // For further informations about the DA working features see:
  20 // Phys.Lett. B601 (2004) 56-63 (http://arxiv.org/abs/hep-ph/0407214)
  21 // Author: Davide Perrino (davide.perrino@ba.infn.it, davide.perrino@cern.ch)
  22 //-----------------------------------------------------------------------------------
  23
  24 #include <TMath.h>
  25 #include <TRandom2.h>
  26 #include <TClonesArray.h>
  27 #include "AliJetReaderHeader.h"
  28 #include "AliJetReader.h"
  29 #include "AliDAJetHeader.h"
  30 #include "AliDAJetFinder.h"
  31
  32 ClassImp(AliDAJetFinder)
  33
  34
  35 //-----------------------------------------------------------------------------------
  36 AliDAJetFinder::AliDAJetFinder():
  37         AliJetFinder(),
  38         fAlpha(1.01),
  39         fDelta(1e-8),
  40         fAvDist(1e-6),
  41         fEps(1e-4),
  42         fEpsMax(1e-2),
  43         fNloopMax(100),
  44         fBeta(0.1),
  45         fNclustMax(0),
  46         fNin(0),
  47         fNeff(0)
  48 {
  49         // Constructor
  50 }
  51
  52 //-----------------------------------------------------------------------------------
  53 AliDAJetFinder::~AliDAJetFinder()
  54 {
  55         // Destructor
  56 }
  57
  58 //-----------------------------------------------------------------------------------
  59 void AliDAJetFinder::FindJets()
  60 {
  61 // Find the jets in current event
  62 //
  63         Float_t betaStop=100.;
  64         fDebug = fHeader->GetDebug();
  65
  66         Double_t dEtSum=0;
  67         Double_t *xData[2];
  68         TVectorD *vPx = new TVectorD();
  69         TVectorD *vPy = new TVectorD();
  70         TMatrixD *mPyx= new TMatrixD();
  71         TMatrixD *mY  = new TMatrixD();
  72         InitDetAnn(dEtSum,xData,vPx,vPy,mPyx,mY);
  73         if (fNin < fNclustMax){
  74           delete [] xData[0], delete [] xData[1];
  75           delete vPx;
  76           delete vPy;
  77           delete mPyx;
  78           delete mY;
  79           return;
  80         }
  81         Int_t nc=1, nk;
  82         DoubleClusters(nc,nk,vPy,mY);
  83         do{                                     //loop over beta
  84                 fBeta*=fAlpha;
  85                 Annealing(nk,xData,vPx,vPy,mPyx,mY);
  86                 NumCl(nc,nk,vPy,mPyx,mY);
  87         }while((fBeta<betaStop || nc<4) && nc<fNclustMax);
  88
  89         Int_t *xx=new Int_t[fNeff];
  90         for (Int_t i = 0; i < fNeff; i++) xx[i] = 0;
  91
  92         EndDetAnn(nk,xData,xx,dEtSum,vPx,vPy,mPyx,mY);
  93         StoreJets(nk,xData,xx,mY);
  94         delete [] xx;
  95
  96         delete [] xData[0], delete [] xData[1];
  97         delete mPyx;
  98         delete mY;
  99         delete vPx;
 100         delete vPy;
 101
 102 }
 103
 104 //-----------------------------------------------------------------------------------
 105 void AliDAJetFinder::InitDetAnn(Double_t &dEtSum,Double_t **xData,TVectorD *vPx,TVectorD *vPy,TMatrixD *mPyx,TMatrixD *mY)
 106 {
 107 //Initialise the variables used by the algorithm
 108         fBeta=0.1;
 109         fNclustMax = ((AliDAJetHeader*)fHeader)->GetFixedCl() ?
 110             ((AliDAJetHeader*)fHeader)->GetNclustMax() :
 111             TMath::Max((Int_t)TMath::Sqrt(fNin),5);
 112         Float_t etaEff = ((AliDAJetHeader*)fHeader)->GetEtaEff();
 113         TClonesArray *lvArray = fReader->GetMomentumArray();
 114         Int_t nEntr = lvArray->GetEntries();
 115         fNin=0;
 116         for (Int_t iEn=0; iEn<nEntr; iEn++) if (fReader->GetCutFlag(iEn)==1) fNin++;
 117
 118         fNeff = ((AliDAJetHeader*)fHeader)->GetNeff();
 119         fNeff = TMath::Max(fNeff,fNin);
 120         Double_t *xEta = new Double_t[fNeff];
 121         Double_t *xPhi = new Double_t[fNeff];
 122         xData[0]=xEta; xData[1]=xPhi;
 123         vPx->ResizeTo(fNeff);
 124         Int_t iIn=0;
 125         for (Int_t iEn=0; iEn<nEntr; iEn++){
 126                 if (fReader->GetCutFlag(iEn)==0) continue;
 127                 TLorentzVector *lv=(TLorentzVector*)lvArray->At(iEn);
 128                 xEta[iIn] = lv->Eta();
 129                 xPhi[iIn] = lv->Phi()<0 ? lv->Phi() + 2*TMath::Pi() : lv->Phi();
 130                 (*vPx)(iIn)=lv->Pt();
 131                 dEtSum+=(*vPx)(iIn);
 132                 iIn++;
 133         }
 134         TRandom2 r;
 135         r.SetSeed(0);
 136         for (iIn=fNin; iIn<fNeff; iIn++){
 137                 xEta[iIn]=r.Uniform(-1*etaEff,etaEff);
 138                 xPhi[iIn]=r.Uniform(0.,2*TMath::Pi());
 139                 (*vPx)(iIn)=r.Uniform(0.01,0.02);
 140                 dEtSum+=(*vPx)(iIn);
 141         }
 142         for (iIn=0; iIn<fNeff; iIn++) (*vPx)(iIn)=(*vPx)(iIn)/dEtSum;
 143
 144         Int_t njdim=2*fNclustMax+1;
 145         mPyx->ResizeTo(fNeff,njdim);
 146         mY->ResizeTo(4,njdim);
 147         vPy->ResizeTo(njdim);
 148         mY->Zero();mPyx->Zero();vPy->Zero();
 149         (*vPy)(0)=1;
 150         TMatrixDColumn(*mPyx,0)=1;
 151         Double_t ypos=0,xpos=0;
 152         for (iIn=0; iIn<fNeff; iIn++){
 153                 (*mY)(0,0)+=(*vPx)(iIn)*xEta[iIn];
 154                 ypos+=(*vPx)(iIn)*TMath::Sin(xPhi[iIn]);
 155                 xpos+=(*vPx)(iIn)*TMath::Cos(xPhi[iIn]);
 156         }
 157         (*mY)(1,0)=(atan2(ypos,xpos)>0) ? atan2(ypos,xpos) : atan2(ypos,xpos)+2*TMath::Pi();
 158 }
 159
 160 //-----------------------------------------------------------------------------------
 161 void AliDAJetFinder::DoubleClusters(Int_t nc,Int_t &nk,  TVectorD *vPy, TMatrixD *mY) const
 162 {
 163 // Return double clusters
 164         for(Int_t iClust=0; iClust<nc; iClust++){
 165                 (*vPy)(iClust)=(*vPy)(iClust)/2;
 166                 (*vPy)(nc+iClust)=(*vPy)(iClust);
 167                 for(Int_t iComp=0; iComp<3; iComp++) (*mY)(iComp,nc+iClust)=(*mY)(iComp,iClust);
 168         }
 169         nk=2*nc;
 170 }
 171
 172 //-----------------------------------------------------------------------------------
 173 void AliDAJetFinder::Annealing(Int_t nk,Double_t **xData,  TVectorD *vPx,  TVectorD *vPy,  TMatrixD *mPyx,  TMatrixD *mY)
 174 {
 175 // Main part of the algorithm
 176         const Double_t pi=TMath::Pi();
 177         TVectorD *py = new TVectorD(nk);
 178         TVectorD *p  = new TVectorD(nk);
 179         TMatrixD *y  = new TMatrixD(4,nk);
 180         TMatrixD *y1 = new TMatrixD(4,nk);
 181         TMatrixD *ry = new TMatrixD(2,nk);
 182         Double_t *xEta = xData[0];
 183         Double_t *xPhi = xData[1];
 184         Double_t Dist(TVectorD,TVectorD);
 185
 186         Double_t df[2]={fReader->GetReaderHeader()->GetFiducialEtaMax(),pi};
 187         TVectorD vPart(2);
 188         Double_t *m = new Double_t[nk];
 189         Double_t chi,chi1;
 190         do{
 191                 Int_t nloop=0;
 192                 for (Int_t iClust=0; iClust<nk; iClust++){
 193                         for (Int_t i=0; i<3; i++)(*y1)(i,iClust)=(*mY)(i,iClust);
 194                         (*py)(iClust)=(*vPy)(iClust);
 195                 }
 196         //perturbation of codevectors
 197                 Double_t seed=1000000*gRandom->Rndm(24);
 198                 ry->Randomize(-0.5,0.5,seed);
 199                 for (Int_t i=0; i<2; i++){
 200                         for (Int_t iClust=0; iClust<nk/2; iClust++)
 201                                 (*y1)(i,iClust)+=((*ry)(i,iClust)+TMath::Sign(0.5,(*ry)(i,iClust)))*fDelta*df[i];
 202                         for (Int_t iClust=nk/2; iClust<nk; iClust++)
 203                                 (*y1)(i,iClust)-=((*ry)(i,iClust-nk/2)+TMath::Sign(0.5,(*ry)(i,iClust-nk/2)))*fDelta*df[i];
 204                 }
 205                 do{
 206         //recalculate conditional probabilities
 207                         nloop++;
 208                         for (Int_t iIn=0; iIn<fNeff; iIn++){
 209                                 vPart(0)=xEta[iIn]; vPart(1)=xPhi[iIn];
 210                                 for(Int_t iClust=0; iClust<nk; iClust++){
 211                                         (*mPyx)(iIn,iClust)=-log((*py)(iClust))+fBeta*Dist(vPart,TMatrixDColumn(*y1,iClust));
 212                                         m[iClust]=(*mPyx)(iIn,iClust);
 213                                 }
 214                                 Double_t pyxNorm=0;
 215                                 Double_t minPyx=TMath::MinElement(nk,m);
 216                                 for (Int_t iClust=0; iClust<nk; iClust++){
 217                                         (*mPyx)(iIn,iClust)-=minPyx;
 218                                         (*mPyx)(iIn,iClust)=exp(-(*mPyx)(iIn,iClust));
 219                                         pyxNorm+=(*mPyx)(iIn,iClust);
 220                                 }
 221                                 for (Int_t iClust=0; iClust<nk; iClust++) (*mPyx)(iIn,iClust)/=pyxNorm;
 222                         }
 223                         p->Zero();
 224                         y->Zero();
 225         //recalculate codevectors
 226                         for (Int_t iClust=0; iClust<nk; iClust++){
 227                                 Double_t xpos=0,ypos=0,pxy;
 228                                 for (Int_t iIn=0; iIn<fNeff; iIn++) (*p)(iClust)+=(*vPx)(iIn)*(*mPyx)(iIn,iClust);
 229                                 for (Int_t iIn=0; iIn<fNeff; iIn++){
 230                                         pxy=(*vPx)(iIn)*(*mPyx)(iIn,iClust)/(*p)(iClust);
 231                                         ypos+=pxy*TMath::Sin(xPhi[iIn]);
 232                                         xpos+=pxy*TMath::Cos(xPhi[iIn]);
 233                                         (*y)(0,iClust)+=pxy*xEta[iIn];
 234                                 }
 235                                 (*y)(1,iClust)=(atan2(ypos,xpos)>0) ? atan2(ypos,xpos) : atan2(ypos,xpos)+2*pi;
 236                         }
 237         //verify codevectors' stability
 238                         chi=0;
 239                         for (Int_t iClust=0; iClust<nk; iClust++){
 240                                 chi1=TMath::CosH((*y1)(0,iClust)-(*y)(0,iClust))-TMath::Cos((*y1)(1,iClust)-(*y)(1,iClust));
 241                                 chi1/=(2*TMath::CosH((*y1)(0,iClust))*TMath::CosH((*y)(0,iClust)));
 242                                 chi1*=chi1;
 243                                 if (chi1>chi) chi=chi1;
 244                         }
 245                         chi=TMath::Sqrt(chi);
 246                         for (Int_t iClust=0; iClust<nk; iClust++){
 247                                 for (Int_t i=0; i<2; i++) (*y1)(i,iClust)=(*y)(i,iClust);
 248                                 (*py)(iClust)=(*p)(iClust);
 249                         }
 250                         if (nloop>fNloopMax){
 251                                 if (chi<fEpsMax || nloop>500) break;
 252                         }
 253                 }while (chi>fEps);
 254         }while (chi>fEpsMax);
 255         for (Int_t iClust=0; iClust<nk; iClust++){                              //set codevectors and probability equal to those calculated
 256                 for (Int_t i=0; i<2; i++) (*mY)(i,iClust)=(*y)(i,iClust);
 257                 (*vPy)(iClust)=(*p)(iClust);
 258         }
 259     delete py;
 260     delete p;
 261     delete y;
 262     delete y1;
 263     delete ry;
 264     delete [] m;
 265 }
 266
 267 //-----------------------------------------------------------------------------------
 268 void AliDAJetFinder::NumCl(Int_t &nc,Int_t &nk,TVectorD *vPy,  TMatrixD *mPyx,TMatrixD *mY)
 269 {
 270     // Number of clusters
 271         static Bool_t growcl=true;
 272
 273         if (nk==2) growcl=true;
 274         if (growcl){
 275 //verify if two codevectors are equal within fAvDist
 276                 Int_t *nSame = new Int_t[nk];
 277                 Int_t **iSame = new Int_t*[nk];
 278                 Int_t **cont = new Int_t*[nk];
 279                 for (Int_t iClust=0; iClust<nk; iClust++) {
 280                     cont[iClust] =new Int_t[nk];
 281                     iSame[iClust]=new Int_t[nk];
 282                 }
 283
 284                 for (Int_t iClust=0; iClust<nk; iClust++){
 285                         iSame[iClust][iClust]=1;
 286                         for (Int_t iClust1=iClust+1; iClust1<nk; iClust1++){
 287                                 Double_t eta  = (*mY)(0,iClust) ; Double_t phi  = (*mY)(1,iClust);
 288                                 Double_t eta1 = (*mY)(0,iClust1); Double_t phi1 = (*mY)(1,iClust1);
 289                                 Double_t distCl=(TMath::CosH(eta-eta1)-TMath::Cos(phi-phi1))/(2*TMath::CosH(eta)*TMath::CosH(eta1));
 290                                 if (distCl < fAvDist) iSame[iClust][iClust1]=iSame[iClust1][iClust]=1;
 291                                 else iSame[iClust][iClust1]=iSame[iClust1][iClust]=0;
 292                         }
 293                 }
 294                 ReduceClusters(iSame,nk,nc,cont,nSame);
 295                 if (nc >= fNclustMax) growcl=false;
 296 //recalculate the nc distinct codevectors
 297                 TMatrixD *pyx = new TMatrixD(fNeff,2*nc);
 298                 TVectorD *py = new TVectorD(nk);
 299                 TMatrixD *y1  = new TMatrixD(3,nk);
 300                 for (Int_t iClust=0; iClust<nc; iClust++){
 301                         for(Int_t j=0; j<nSame[iClust]; j++){
 302                                 Int_t iClust1 = cont[iClust][j];
 303                                 for (Int_t iIn=0; iIn<fNeff; iIn++) (*pyx)(iIn,iClust)+=(*mPyx)(iIn,iClust1);
 304                                 (*py)(iClust)+=(*vPy)(iClust1);
 305                                 for (Int_t i=0; i<2; i++) (*y1)(i,iClust)+=(*mY)(i,iClust1);
 306                         }
 307                         for (Int_t i=0; i<2; i++) (*y1)(i,iClust)/=nSame[iClust];
 308                 }
 309                 for (Int_t iClust=0; iClust<nk; iClust++) delete [] cont[iClust], delete [] iSame[iClust];
 310                 delete [] iSame;
 311                 delete [] cont;
 312                 delete [] nSame;
 313                 if (nc > nk/2){
 314                         for (Int_t iClust=0; iClust<nc; iClust++){
 315                                 for (Int_t iIn=0; iIn<fNeff; iIn++) (*mPyx)(iIn,iClust)=(*pyx)(iIn,iClust);
 316                                 for (Int_t iComp=0; iComp<2; iComp++) (*mY)(iComp,iClust)=(*y1)(iComp,iClust);
 317                                 (*vPy)(iClust)=(*py)(iClust);
 318                         }
 319                         nk=nc;
 320                         if (growcl) DoubleClusters(nc,nk,vPy,mY);
 321                 }
 322                 delete pyx;
 323                 delete py;
 324                 delete y1;
 325         }
 326
 327 }
 328
 329 //-----------------------------------------------------------------------------------
 330 void AliDAJetFinder::ReduceClusters(Int_t **iSame,Int_t nc,Int_t &ncout,Int_t **cont,Int_t *nSameOut) const
 331 {
 332 // Reduction step
 333         Int_t *nSame = new Int_t[nc];
 334         Int_t *iperm = new Int_t[nc];
 335         Int_t *go = new Int_t[nc];
 336         for (Int_t iCl=0; iCl<nc; iCl++){
 337                 nSame[iCl]=0;
 338                 for (Int_t jCl=0; jCl<nc; jCl++) nSame[iCl]+=iSame[iCl][jCl], cont[iCl][jCl]=0;
 339                 iperm[iCl]=iCl;
 340                 go[iCl]=1;
 341         }
 342         TMath::Sort(nc,nSame,iperm,true);
 343         Int_t l=0;
 344         for (Int_t iCl=0; iCl<nc; iCl++){
 345                 Int_t k=iperm[iCl];
 346                 if (go[k] == 1){
 347                         Int_t m=0;
 348                         for (Int_t jCl=0; jCl<nc; jCl++){
 349                                 if (iSame[k][jCl] == 1){
 350                                         cont[l][m]=jCl;
 351                                         go[jCl]=0;
 352                                         m++;
 353                                 }
 354                         }
 355                         nSameOut[l]=m;
 356                         l++;
 357                 }
 358         }
 359         ncout=l;
 360         delete [] nSame;
 361         delete [] iperm;
 362         delete [] go;
 363 }
 364
 365 //-----------------------------------------------------------------------------------
 366 void AliDAJetFinder::EndDetAnn(Int_t &nk,Double_t **xData,Int_t *xx,Double_t etx,TVectorD *vPx,TVectorD *vPy,TMatrixD *mPyx,TMatrixD *mY)
 367 {
 368 //now assign each particle to only one cluster
 369         Double_t *clusters=new Double_t[nk];
 370         for (Int_t iIn=0; iIn<fNeff; iIn++){
 371                 for (Int_t iClust=0; iClust<nk; iClust++) clusters[iClust]=(*mPyx)(iIn,iClust);
 372                 xx[iIn]=TMath::LocMax(nk,clusters);
 373         }
 374         delete [] clusters;
 375
 376 //recalculate codevectors, having all p(y|x)=0 or 1
 377         Double_t *xEta = xData[0];
 378         Double_t *xPhi = xData[1];
 379         mY->Zero();
 380         mPyx->Zero();
 381         vPy->Zero();
 382         for (Int_t iIn=0; iIn<fNin; iIn++){
 383                 Int_t iClust=xx[iIn];
 384                 (*mPyx)(iIn,iClust)=1;
 385                 (*vPy)(iClust)+=(*vPx)(iIn);
 386                 (*mY)(0,iClust)+=(*vPx)(iIn)*xEta[iIn];
 387                 (*mY)(3,iClust)+=(*vPx)(iIn)*etx;
 388         }
 389         Int_t k=0;
 390         for (Int_t iClust=0; iClust<nk; iClust++){
 391                 if ((*vPy)(iClust)>0){
 392                         Double_t xpos=0,ypos=0,pxy;
 393                         for (Int_t iIn=0; iIn<fNin; iIn++){
 394                                 pxy=(*vPx)(iIn)*(*mPyx)(iIn,iClust)/(*vPy)(iClust);
 395                                 ypos+=pxy*TMath::Sin(xPhi[iIn]);
 396                                 xpos+=pxy*TMath::Cos(xPhi[iIn]);
 397                                 if (xx[iIn]==iClust) xx[iIn]=k;
 398                         }
 399                         (*mY)(0,k)=(*mY)(0,iClust)/(*vPy)(iClust);
 400                         (*mY)(1,k)=(atan2(ypos,xpos)>0) ? atan2(ypos,xpos) : atan2(ypos,xpos)+2*TMath::Pi();
 401                         (*mY)(3,k)=(*mY)(3,iClust);
 402                         k++;
 403                 }
 404         }
 405         nk=k;
 406 }
 407
 408 //-----------------------------------------------------------------------------------
 409 void AliDAJetFinder::StoreJets(Int_t nk, Double_t **xData, Int_t *xx, TMatrixD *mY)
 410 {
 411 //evaluate significant clusters properties
 412         const Double_t pi=TMath::Pi();
 413         AliJetReaderHeader *rHeader=fReader->GetReaderHeader();
 414         Float_t dFidEtaMax = rHeader->GetFiducialEtaMax();
 415         Float_t dFidEtaMin = rHeader->GetFiducialEtaMin();
 416         Float_t dFiducialEta= dFidEtaMax - dFidEtaMin;
 417         Double_t *xEta = xData[0];
 418         Double_t *xPhi = xData[1];
 419         Int_t nEff = 0;
 420         for (Int_t i=0; i<fNeff; i++) if (xEta[i]<dFidEtaMax && xEta[i]>dFidEtaMin) nEff++;
 421         Double_t dMeanDist=0.;
 422         if (nEff > 0)
 423             dMeanDist=TMath::Sqrt(2*dFiducialEta*pi/nEff);
 424         Bool_t   *isJet = new Bool_t[nk];
 425         Double_t *etNoBg= new Double_t[nk];
 426         Double_t *dDeltaEta=new Double_t[nk];
 427         Double_t *dDeltaPhi=new Double_t[nk];
 428         Double_t *surf  = new Double_t[nk];
 429         Double_t *etDens= new Double_t[nk];
 430         for (Int_t iClust=0; iClust<nk; iClust++){
 431                 isJet[iClust]=false;
 432                 Double_t dEtaMin=10.,dEtaMax=-10.,dPhiMin=10.,dPhiMax=0.;
 433                 for (Int_t iIn=0; iIn<fNeff; iIn++){
 434                         if (xx[iIn]!=iClust || xEta[iIn]>dFidEtaMax || xEta[iIn]<dFidEtaMin) continue;
 435                         if (xEta[iIn] < dEtaMin) dEtaMin=xEta[iIn];
 436                         if (xEta[iIn] > dEtaMax) dEtaMax=xEta[iIn];
 437                         Double_t dPhi=xPhi[iIn]-(*mY)(1,iClust);
 438                         if      (dPhi > pi     ) dPhi-=2*pi;
 439                         else if (dPhi < (-1)*pi) dPhi+=2*pi;
 440                         if      (dPhi < dPhiMin) dPhiMin=dPhi;
 441                         else if (dPhi > dPhiMax) dPhiMax=dPhi;
 442                 }
 443                 dDeltaEta[iClust]=dEtaMax-dEtaMin+dMeanDist;
 444                 dDeltaPhi[iClust]=dPhiMax-dPhiMin+dMeanDist;
 445                 surf[iClust]=0.25*pi*dDeltaEta[iClust]*dDeltaPhi[iClust];
 446                 etDens[iClust]=(*mY)(3,iClust)/surf[iClust];
 447         }
 448
 449         if (((AliDAJetHeader*)fHeader)->GetSelJets()){
 450                 for (Int_t iClust=0; iClust<nk; iClust++){
 451                         if (!isJet[iClust] && (*mY)(0,iClust)<dFidEtaMax && (*mY)(0,iClust)>dFidEtaMin){
 452                                 Double_t etDensMed=0.;
 453                                 Double_t etDensSqr=0.;
 454                                 Int_t norm=0;
 455                                 for (Int_t iClust1=0; iClust1<nk; iClust1++){
 456                                         if(iClust1!=iClust && (*mY)(0,iClust)<dFidEtaMax && (*mY)(0,iClust)>dFidEtaMin){
 457                                                 etDensMed+=etDens[iClust1];
 458                                                 etDensSqr+=TMath::Power(etDens[iClust1],2);
 459                                                 norm++;
 460                                         }
 461                                 }
 462                                 etDensMed/=TMath::Max(norm,1);
 463                                 etDensSqr/=TMath::Max(norm,1);
 464                                 Double_t deltaEtDens=TMath::Sqrt(etDensSqr-TMath::Power(etDensMed,2));
 465                                 if ((*mY)(3,iClust) > (etDensMed+deltaEtDens)*surf[iClust]) isJet[iClust]=kTRUE;
 466                                 etNoBg[iClust]=(*mY)(3,iClust)-etDensMed*surf[iClust];
 467                         }
 468                 }
 469                 for (Int_t iClust=0; iClust<nk; iClust++){
 470                         if (isJet[iClust]){
 471                                 Double_t etDensMed=0;
 472                                 Double_t extSurf=2*dFiducialEta*pi;
 473                                 for (Int_t iClust1=0; iClust1<nk; iClust1++){
 474                                         if (!isJet[iClust1]) etDensMed+=(*mY)(3,iClust1);
 475                                         else extSurf-=surf[iClust1];
 476                                 }
 477                                 etDensMed/=extSurf;
 478                                 etNoBg[iClust]=(*mY)(3,iClust)-etDensMed*surf[iClust];
 479                                 if (etNoBg[iClust]<((AliDAJetHeader*)fHeader)->GetEtMin()){
 480                                         isJet[iClust]=kFALSE;
 481                                         iClust=-1;
 482                                 }
 483                         }
 484                 }
 485         } else {
 486                 for (Int_t iClust=0; iClust<nk; iClust++){
 487                         isJet[iClust]=true;
 488                         etNoBg[iClust]=(*mY)(3,iClust);
 489                 }
 490         }
 491         delete [] etDens;
 492         delete [] surf;
 493
 494 //now add selected jets to the list
 495         Int_t *iSort = new Int_t[nk];
 496         TMath::Sort(nk,etNoBg,iSort,true);
 497         Int_t iCl = 0;
 498         TRefArray *refs = 0;
 499         Bool_t fromAod = !strcmp(fReader->ClassName(),"AliJetAODReader");
 500         if (fromAod) refs = fReader->GetReferences();
 501         for (Int_t iClust=0; iClust<nk; iClust++){                                                                      //clusters loop
 502                 iCl=iSort[iClust];
 503                 if (isJet[iCl]){                                                                                                                //choose cluster
 504                         Float_t px,py,pz,en;
 505                         px = (*mY)(3,iCl)*TMath::Cos((*mY)(1,iCl));
 506                         py = (*mY)(3,iCl)*TMath::Sin((*mY)(1,iCl));
 507                         pz = (*mY)(3,iCl)/TMath::Tan(2.0 * TMath::ATan(TMath::Exp(-(*mY)(0,iCl))));
 508                         en = TMath::Sqrt(px * px + py * py + pz * pz);
 509                         AliAODJet jet(px, py, pz, en);
 510                         if (fromAod){
 511                                 Int_t iIn=0;
 512                                 Int_t nEntr = fReader->GetMomentumArray()->GetEntries();
 513                                 for (Int_t iEn=0; iEn<nEntr; iEn++){
 514                                         if (fReader->GetCutFlag(iEn)==0) continue;
 515                                         if (xx[iIn]==iCl) jet.AddTrack(refs->At(iEn));
 516                                         iIn++;
 517                                 }
 518                         }
 519                         AddJet(jet);
 520                         if (fDebug > 0) printf("jet %d, Eta: %f, Phi: %f, Et: %f\n",iCl,jet.Eta(),jet.Phi(),jet.Pt());
 521                 }
 522         }
 523         delete [] dDeltaEta; delete [] dDeltaPhi;
 524         delete [] etNoBg;
 525         delete [] isJet;
 526         delete [] iSort;
 527 }
 528
 529 //-----------------------------------------------------------------------------------
 530 Double_t Dist(TVectorD x,TVectorD y)
 531 {
 532 // Squared distance
 533         const Double_t pi=TMath::Pi();
 534         Double_t dphi=TMath::Abs(x(1)-y(1));
 535         if (dphi > pi) dphi=2*pi-dphi;
 536         Double_t dist=pow(x(0)-y(0),2)+pow(dphi,2);
 537         return dist;
 538 }