[u/mrichter/AliRoot.git] / JETAN / AliDAJetFinder.cxx


//  **************************************************************************
//  * Copyright(c) 1998-1999, ALICE Experiment at CERN, All rights reserved. *
//  *                                                                        *
//  * Author: The ALICE Off-line Project.                                    *
//  * Contributors are mentioned in the code where appropriate.              *
//  *                                                                        *
//  * Permission to use, copy, modify and distribute this software and its   *
//  * documentation strictly for non-commercial purposes is hereby granted   *
//  * without fee, provided that the above copyright notice appears in all   *
//  * copies and that both the copyright notice and this permission notice   *
//  * appear in the supporting documentation. The authors make no claims     *
//  * about the suitability of this software for any purpose. It is          *
//  * provided "as is" without express or implied warranty.                  *
//  **************************************************************************

//-----------------------------------------------------------------------------------
// Jet finder based on Deterministic Annealing
// For further informations about the DA working features see:
// Phys.Lett. B601 (2004) 56-63 (http://arxiv.org/abs/hep-ph/0407214)
// Author: Davide Perrino (davide.perrino@ba.infn.it, davide.perrino@cern.ch)
//-----------------------------------------------------------------------------------

#include <TMath.h>
#include <TRandom.h>
#include <TClonesArray.h>
#include "AliJetReader.h"
#include "AliDAJetHeader.h"
#include "AliDAJetFinder.h"


ClassImp(AliDAJetFinder)


//-----------------------------------------------------------------------------------
AliDAJetFinder::AliDAJetFinder():
	fAlpha(1.01),
	fDelta(1e-8),
	fAvDist(1e-6),
	fEps(1e-4),
	fEpsMax(1e-2),
	fNloopMax(100),
	fBeta(0.1),
	fNclustMax(0),
	fPyx(0x0),
	fY(0x0),
	fPx(0x0),
	fPy(0x0),
	fXEta(0x0),
	fXPhi(0x0),
	fNin(0)
{
	// Constructor
}

//-----------------------------------------------------------------------------------
AliDAJetFinder::~AliDAJetFinder()
{
	// Destructor
	delete fPyx;
	delete fY;
	delete fPx;
	delete fPy;
	delete [] fXEta;
	delete [] fXPhi;
}

//-----------------------------------------------------------------------------------
void AliDAJetFinder::FindJets()  
{
// Find the jets in current event
// 
	Float_t betaStop=100.;

	Double_t dEtSum=0;
	InitDetAnn(dEtSum);
	if (!fNin) return;

	Int_t nc=1,nk;
	DoubleClusters(nc,nk);
	do{					//loop over beta
		fBeta*=fAlpha;
		Annealing(nk);
		NumCl(nc,nk);
	}while((fBeta<betaStop || nc<4) && nc<fNclustMax);

	Int_t *xx=new Int_t[fNin];
	EndDetAnn(nk,xx,dEtSum);
	StoreJets(nk,xx);
	delete [] xx;

}

//-----------------------------------------------------------------------------------
void AliDAJetFinder::InitDetAnn(Double_t &dEtSum)
{
//Initialise the variables used by the algorithm
	fBeta=0.1;
	TClonesArray *lvArray = fReader->GetMomentumArray();
	fNin = lvArray->GetEntries();
	fNclustMax= ((AliDAJetHeader*)fHeader)->GetFixedCl() ? 
	    ((AliDAJetHeader*)fHeader)->GetNclustMax() 
	    : 
	    TMath::Max((Int_t)TMath::Sqrt(fNin),5);
	fXEta=new Double_t[fNin]; fXPhi=new Double_t[fNin];
	fPx = new TVectorD(fNin);
	for (Int_t iIn=0; iIn<fNin; iIn++){
		TLorentzVector *lv=(TLorentzVector*)lvArray->At(iIn);
		fXEta[iIn] = lv->Eta();
		fXPhi[iIn] = lv->Phi()<0 ? lv->Phi() + 2*TMath::Pi() : lv->Phi();
		(*fPx)(iIn)=lv->Pt();
		dEtSum+=(*fPx)(iIn);
	}
	for (Int_t iIn=0; iIn<fNin; iIn++) (*fPx)(iIn)=(*fPx)(iIn)/dEtSum;

	Int_t njdim=2*fNclustMax+1;
	fPyx = new TMatrixD(fNin,njdim);
	fY = new TMatrixD(4,njdim);
	fPy= new TVectorD(njdim);
	fY->Zero();fPyx->Zero();fPy->Zero();
	(*fPy)(0)=1;
	TMatrixDColumn(*fPyx,0)=1;
	Double_t ypos=0,xpos=0;
	for (Int_t iIn=0; iIn<fNin; iIn++){
		(*fY)(0,0)+=(*fPx)(iIn)*fXEta[iIn];
		ypos+=(*fPx)(iIn)*TMath::Sin(fXPhi[iIn]);
		xpos+=(*fPx)(iIn)*TMath::Cos(fXPhi[iIn]);
	}
	(*fY)(1,0)=(atan2(ypos,xpos)>0) ? atan2(ypos,xpos) : atan2(ypos,xpos)+2*TMath::Pi();
	lvArray->Delete();
}

//-----------------------------------------------------------------------------------
void AliDAJetFinder::DoubleClusters(Int_t nc,Int_t &nk)
{
	for(Int_t iClust=0; iClust<nc; iClust++){
		(*fPy)(iClust)=(*fPy)(iClust)/2;
		(*fPy)(nc+iClust)=(*fPy)(iClust);
		for(Int_t iComp=0; iComp<3; iComp++) (*fY)(iComp,nc+iClust)=(*fY)(iComp,iClust);
	}
	nk=2*nc;
}

//-----------------------------------------------------------------------------------
void AliDAJetFinder::Annealing(Int_t nk)
{
// Main part of the algorithm
	const Double_t pi=TMath::Pi();
	TVectorD *py = new TVectorD(nk);
	TVectorD *p  = new TVectorD(nk);
	TMatrixD *y  = new TMatrixD(4,nk);
	TMatrixD *y1 = new TMatrixD(4,nk);
	TMatrixD *ry = new TMatrixD(2,nk);
	Double_t Dist(TVectorD,TVectorD);

	Double_t df[2]={((AliDAJetHeader*)fHeader)->GetEtaCut(),pi};
	TVectorD vPart(2);
	Double_t *m = new Double_t[nk];
	Double_t chi,chi1;
	do{
		Int_t nloop=0;
		for (Int_t iClust=0; iClust<nk; iClust++){
			for (Int_t i=0; i<3; i++)(*y1)(i,iClust)=(*fY)(i,iClust);
			(*py)(iClust)=(*fPy)(iClust);
		}
	//perturbation of codevectors
		Double_t seed=1000000*gRandom->Rndm(24);
		ry->Randomize(-0.5,0.5,seed);
		for (Int_t i=0; i<2; i++){
			for (Int_t iClust=0; iClust<nk/2; iClust++)
				(*y1)(i,iClust)+=((*ry)(i,iClust)+TMath::Sign(0.5,(*ry)(i,iClust)))*fDelta*df[i];
			for (Int_t iClust=nk/2; iClust<nk; iClust++)
				(*y1)(i,iClust)-=((*ry)(i,iClust-nk/2)+TMath::Sign(0.5,(*ry)(i,iClust-nk/2)))*fDelta*df[i];
		}
		do{
	//recalculate conditional probabilities
			nloop++;
			for (Int_t iIn=0; iIn<fNin; iIn++){
				vPart(0)=fXEta[iIn]; vPart(1)=fXPhi[iIn];
				for(Int_t iClust=0; iClust<nk; iClust++){
					(*fPyx)(iIn,iClust)=-log((*py)(iClust))+fBeta*Dist(vPart,TMatrixDColumn(*y1,iClust));
					m[iClust]=(*fPyx)(iIn,iClust);
				}
				Double_t pyxNorm=0;
				Double_t minPyx=TMath::MinElement(nk,m);
				for (Int_t iClust=0; iClust<nk; iClust++){
					(*fPyx)(iIn,iClust)-=minPyx;
					(*fPyx)(iIn,iClust)=exp(-(*fPyx)(iIn,iClust));
					pyxNorm+=(*fPyx)(iIn,iClust);
				}
				for (Int_t iClust=0; iClust<nk; iClust++) (*fPyx)(iIn,iClust)/=pyxNorm;
			}
			p->Zero();
			y->Zero();
	//recalculate codevectors
			for (Int_t iClust=0; iClust<nk; iClust++){
				Double_t xpos=0,ypos=0,pxy;
				for (Int_t iIn=0; iIn<fNin; iIn++) (*p)(iClust)+=(*fPx)(iIn)*(*fPyx)(iIn,iClust);
				for (Int_t iIn=0; iIn<fNin; iIn++){
					pxy=(*fPx)(iIn)*(*fPyx)(iIn,iClust)/(*p)(iClust);
					ypos+=pxy*TMath::Sin(fXPhi[iIn]);
					xpos+=pxy*TMath::Cos(fXPhi[iIn]);
					(*y)(0,iClust)+=pxy*fXEta[iIn];
				}
				(*y)(1,iClust)=(atan2(ypos,xpos)>0) ? atan2(ypos,xpos) : atan2(ypos,xpos)+2*pi;
			}
	//verify codevectors' stability
			chi=0;
			for (Int_t iClust=0; iClust<nk; iClust++){
				chi1=TMath::CosH((*y1)(0,iClust)-(*y)(0,iClust))-TMath::Cos((*y1)(1,iClust)-(*y)(1,iClust));
				chi1/=(2*TMath::CosH((*y1)(0,iClust))*TMath::CosH((*y)(0,iClust)));
				chi1*=chi1;
				if (chi1>chi) chi=chi1;
			}
			chi=TMath::Sqrt(chi);
			for (Int_t iClust=0; iClust<nk; iClust++){
				for (Int_t i=0; i<2; i++) (*y1)(i,iClust)=(*y)(i,iClust);
				(*py)(iClust)=(*p)(iClust);
			}
			if (nloop>fNloopMax){
				if (chi<fEpsMax || nloop>500) break;
			}
		}while (chi>fEps);
	}while (chi>fEpsMax);
	for (Int_t iClust=0; iClust<nk; iClust++){				//set codevectors and probability equal to those calculated
		for (Int_t i=0; i<2; i++) (*fY)(i,iClust)=(*y)(i,iClust);
		(*fPy)(iClust)=(*p)(iClust);
	}
    delete py;
    delete p;
    delete y;
    delete y1;
    delete ry;
	delete [] m;
}

//-----------------------------------------------------------------------------------
void AliDAJetFinder::NumCl(Int_t &nc,Int_t &nk)
{
	static Bool_t growcl=true;
	
	if (nk==2) growcl=true;
	if (growcl){
//verify if two codevectors are equal within fAvDist
		Int_t *nSame = new Int_t[nk];
		Int_t **iSame = new Int_t*[nk];
		Int_t **cont = new Int_t*[nk];
		for (Int_t iClust=0; iClust<nk; iClust++) cont[iClust]=new Int_t[nk],iSame[iClust]=new Int_t[nk];
		for (Int_t iClust=0; iClust<nk; iClust++){
			iSame[iClust][iClust]=1;
			for (Int_t iClust1=iClust+1; iClust1<nk; iClust1++){
				Double_t eta  = (*fY)(0,iClust) ; Double_t phi  = (*fY)(1,iClust);
				Double_t eta1 = (*fY)(0,iClust1); Double_t phi1 = (*fY)(1,iClust1);
				Double_t distCl=(TMath::CosH(eta-eta1)-TMath::Cos(phi-phi1))/(2*TMath::CosH(eta)*TMath::CosH(eta1));
				if (distCl < fAvDist) iSame[iClust][iClust1]=iSame[iClust1][iClust]=1;
				else iSame[iClust][iClust1]=iSame[iClust1][iClust]=0;
			}
		}
		ReduceClusters(iSame,nk,nc,cont,nSame);
		if (nc >= fNclustMax) growcl=false;
//recalculate the nc distinct codevectors
		TMatrixD *pyx = new TMatrixD(fNin,2*nc);
		TVectorD *py = new TVectorD(nk);
		TMatrixD *y1  = new TMatrixD(3,nk);
		for (Int_t iClust=0; iClust<nc; iClust++){
			for(Int_t j=0; j<nSame[iClust]; j++){
				Int_t iClust1 = cont[iClust][j];
				for (Int_t iIn=0; iIn<fNin; iIn++) (*pyx)(iIn,iClust)+=(*fPyx)(iIn,iClust1);
				(*py)(iClust)+=(*fPy)(iClust1);
				for (Int_t i=0; i<2; i++) (*y1)(i,iClust)+=(*fY)(i,iClust1);
			}
			for (Int_t i=0; i<2; i++) (*y1)(i,iClust)/=nSame[iClust];
		}
		for (Int_t iClust=0; iClust<nk; iClust++) delete [] cont[iClust], delete [] iSame[iClust];
		delete [] iSame;
		delete [] cont;
		delete [] nSame;
		if (nc > nk/2){
			for (Int_t iClust=0; iClust<nc; iClust++){
				for (Int_t iIn=0; iIn<fNin; iIn++) (*fPyx)(iIn,iClust)=(*pyx)(iIn,iClust);
				for (Int_t iComp=0; iComp<2; iComp++) (*fY)(iComp,iClust)=(*y1)(iComp,iClust);
				(*fPy)(iClust)=(*py)(iClust);
			}
			nk=nc;
			if (growcl) DoubleClusters(nc,nk);
		}
		delete pyx;
		delete py;
		delete y1;
	}

}

//-----------------------------------------------------------------------------------
void AliDAJetFinder::ReduceClusters(Int_t **iSame,Int_t nc,Int_t &ncout,Int_t **cont,Int_t *nSameOut)
{
	Int_t *nSame = new Int_t[nc];
	Int_t *iperm = new Int_t[nc];
	Int_t *go = new Int_t[nc];
	for (Int_t iCl=0; iCl<nc; iCl++){
		nSame[iCl]=0;
		for (Int_t jCl=0; jCl<nc; jCl++) nSame[iCl]+=iSame[iCl][jCl], cont[iCl][jCl]=0;
		iperm[iCl]=iCl;
		go[iCl]=1;
	}
	TMath::Sort(nc,nSame,iperm,true);
	Int_t l=0;
	for (Int_t iCl=0; iCl<nc; iCl++){
		Int_t k=iperm[iCl];
		if (go[k] == 1){
			Int_t m=0;
			for (Int_t jCl=0; jCl<nc; jCl++){
				if (iSame[k][jCl] == 1){
					cont[l][m]=jCl;
					go[jCl]=0;
					m++;
				}
			}
			nSameOut[l]=m;
			l++;
		}
	}
	ncout=l;
	delete [] nSame;
	delete [] iperm;
	delete [] go;
}

//-----------------------------------------------------------------------------------
void AliDAJetFinder::EndDetAnn(Int_t &nk,Int_t *xx,Double_t etx)
{
//now assign each particle to only one cluster
	Double_t *clusters=new Double_t[nk];
	for (Int_t iIn=0; iIn<fNin; iIn++){
		for (Int_t iClust=0; iClust<nk; iClust++) clusters[iClust]=(*fPyx)(iIn,iClust);
		xx[iIn]=TMath::LocMax(nk,clusters);
	}
    delete [] clusters;
	
//recalculate codevectors, having all p(y|x)=0 or 1
	fY->Zero();
	fPyx->Zero();
	fPy->Zero();
	for (Int_t iIn=0; iIn<fNin; iIn++){
		Int_t iClust=xx[iIn];
		(*fPyx)(iIn,iClust)=1;
		(*fPy)(iClust)+=(*fPx)(iIn);
		(*fY)(0,iClust)+=(*fPx)(iIn)*fXEta[iIn];
		(*fY)(3,iClust)+=(*fPx)(iIn)*etx;
	}
	Int_t k=0;
	for (Int_t iClust=0; iClust<nk; iClust++){
		if ((*fPy)(iClust)>0){
			Double_t xpos=0,ypos=0,pxy;
			for (Int_t iIn=0; iIn<fNin; iIn++){
				pxy=(*fPx)(iIn)*(*fPyx)(iIn,iClust)/(*fPy)(iClust);
				ypos+=pxy*TMath::Sin(fXPhi[iIn]);
				xpos+=pxy*TMath::Cos(fXPhi[iIn]);
				if (xx[iIn]==iClust) xx[iIn]=k;
			}
			(*fY)(0,k)=(*fY)(0,iClust)/(*fPy)(iClust);
			(*fY)(1,k)=(atan2(ypos,xpos)>0) ? atan2(ypos,xpos) : atan2(ypos,xpos)+2*TMath::Pi();
			(*fY)(3,k)=(*fY)(3,iClust);
			k++;
		}
	}
	nk=k;
}

//-----------------------------------------------------------------------------------
void AliDAJetFinder::StoreJets(Int_t nk,Int_t *xx)
{
//evaluate significant clusters properties
	const Double_t pi=TMath::Pi();
	Double_t dMeanDist=TMath::Sqrt(4*((AliDAJetHeader*)fHeader)->GetEtaCut()*pi/fNin);
	Bool_t   *isJet = new Bool_t[nk];
	Double_t *etNoBg= new Double_t[nk];
	Double_t *dDeltaEta=new Double_t[nk];
	Double_t *dDeltaPhi=new Double_t[nk];
	Double_t *surf  = new Double_t[nk];
	Double_t *etDens= new Double_t[nk];
	for (Int_t iClust=0; iClust<nk; iClust++){									//clusters loop
		isJet[iClust]=false;
		Double_t dEtaMin=10.,dEtaMax=-10.,dPhiMin=10.,dPhiMax=0.;
		for (Int_t iIn=0; iIn<fNin; iIn++){
			if (xx[iIn]!=iClust) continue;
			if (fXEta[iIn] < dEtaMin) dEtaMin=fXEta[iIn];
			if (fXEta[iIn] > dEtaMax) dEtaMax=fXEta[iIn];
			Double_t dPhi=fXPhi[iIn]-(*fY)(1,iClust);
			if      (dPhi > pi     ) dPhi-=2*pi;
			else if (dPhi < (-1)*pi) dPhi+=2*pi;
			if      (dPhi < dPhiMin) dPhiMin=dPhi;
			else if (dPhi > dPhiMax) dPhiMax=dPhi;
		}
		dDeltaEta[iClust]=dEtaMax-dEtaMin+dMeanDist;
		dDeltaPhi[iClust]=dPhiMax-dPhiMin+dMeanDist;
		surf[iClust]=0.25*pi*dDeltaEta[iClust]*dDeltaPhi[iClust];
		etDens[iClust]=(*fY)(3,iClust)/surf[iClust];
	}

	if (((AliDAJetHeader*)fHeader)->GetSelJets()){
		for (Int_t iClust=0; iClust<nk; iClust++){
			if (!isJet[iClust]){
				Double_t etDensMed=0.;
				Double_t etDensSqr=0.;
				Int_t norm=0;
				for (Int_t iClust1=0; iClust1<nk; iClust1++){
					if(iClust1!=iClust){
						etDensMed+=etDens[iClust1];
						etDensSqr+=TMath::Power(etDens[iClust1],2);
						norm++;
					}
				}
				etDensMed/=TMath::Max(norm,1);
				etDensSqr/=TMath::Max(norm,1);
				Double_t deltaEtDens=TMath::Sqrt(etDensSqr-TMath::Power(etDensMed,2));
				if ((*fY)(3,iClust) > (etDensMed+deltaEtDens)*surf[iClust]) isJet[iClust]=kTRUE;
				etNoBg[iClust]=(*fY)(3,iClust)-etDensMed*surf[iClust];
			}
		}
		for (Int_t iClust=0; iClust<nk; iClust++){
			if (isJet[iClust]){
				Double_t etDensMed=0;
				Double_t extSurf=4*((AliDAJetHeader*)fHeader)->GetEtaCut()*pi;
				for (Int_t iClust1=0; iClust1<nk; iClust1++){
					if (!isJet[iClust1]) etDensMed+=(*fY)(3,iClust1);
					else extSurf-=surf[iClust1];
				}
				etDensMed/=extSurf;
				etNoBg[iClust]=(*fY)(3,iClust)-etDensMed*surf[iClust];
				if (etNoBg[iClust]<((AliDAJetHeader*)fHeader)->GetEtMin()){
					isJet[iClust]=kFALSE;
					iClust=-1;
				}
			}
		}
	} else {
		for (Int_t iClust=0; iClust<nk; iClust++) isJet[iClust]=true;
	}
	delete [] etDens;
	delete [] surf;
	
//now add selected jets to the list
	Int_t *inJet = new Int_t[fNin];
	TRefArray *refs = 0;
	Bool_t fromAod = !strcmp(fReader->ClassName(),"AliJetAODReader");
	if (fromAod) refs = fReader->GetReferences();
	for (Int_t iClust=0; iClust<nk; iClust++){									//clusters loop
		if (isJet[iClust]){														//choose cluster
			Float_t px,py,pz,en;
			px = (*fY)(3,iClust)*TMath::Cos((*fY)(1,iClust));
			py = (*fY)(3,iClust)*TMath::Sin((*fY)(1,iClust));
			pz = (*fY)(3,iClust)/TMath::Tan(2.0 * TMath::ATan(TMath::Exp(-(*fY)(0,iClust))));
			en = TMath::Sqrt(px * px + py * py + pz * pz);
			AliAODJet jet(px, py, pz, en);
			if (fromAod) 
			    for (Int_t iIn=0; iIn<fNin; iIn++) if (xx[iIn]==iClust) jet.AddTrack(refs->At(iIn));
			AddJet(jet);
			printf("jet %d, Eta: %f, Phi: %f, Et: %f\n",iClust,jet.Eta(),jet.Phi(),jet.Pt());
		}
	}
	delete [] dDeltaEta; delete [] dDeltaPhi;
	delete [] etNoBg;
	delete [] isJet;
	delete [] inJet;
}

//-----------------------------------------------------------------------------------
Double_t Dist(TVectorD x,TVectorD y)
{
// Squared distance
	const Double_t pi=TMath::Pi();
	Double_t dphi=TMath::Abs(x(1)-y(1));
	if (dphi > pi) dphi=2*pi-dphi;
	Double_t dist=pow(x(0)-y(0),2)+pow(dphi,2);
	return dist;
}
Commit	Line	Data
9e4cc50d	1
7c679be0	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 <TRandom.h>
	26	#include <TClonesArray.h>
	27	#include "AliJetReader.h"
	28	#include "AliDAJetHeader.h"
	29	#include "AliDAJetFinder.h"
	30
	31
	32	ClassImp(AliDAJetFinder)
	33
	34
	35	//-----------------------------------------------------------------------------------
	36	AliDAJetFinder::AliDAJetFinder():
	37	fAlpha(1.01),
	38	fDelta(1e-8),
	39	fAvDist(1e-6),
	40	fEps(1e-4),
	41	fEpsMax(1e-2),
	42	fNloopMax(100),
	43	fBeta(0.1),
	44	fNclustMax(0),
	45	fPyx(0x0),
	46	fY(0x0),
	47	fPx(0x0),
	48	fPy(0x0),
	49	fXEta(0x0),
	50	fXPhi(0x0),
9e4cc50d	51	fNin(0)
7c679be0	52	{
	53	// Constructor
	54	}
	55
	56	//-----------------------------------------------------------------------------------
	57	AliDAJetFinder::~AliDAJetFinder()
	58	{
	59	// Destructor
7c679be0	60	delete fPyx;
	61	delete fY;
	62	delete fPx;
	63	delete fPy;
	64	delete [] fXEta;
	65	delete [] fXPhi;
	66	}
	67
	68	//-----------------------------------------------------------------------------------
	69	void AliDAJetFinder::FindJets()
	70	{
	71	// Find the jets in current event
	72	//
	73	Float_t betaStop=100.;
	74
	75	Double_t dEtSum=0;
	76	InitDetAnn(dEtSum);
	77	if (!fNin) return;
	78
	79	Int_t nc=1,nk;
	80	DoubleClusters(nc,nk);
	81	do{ //loop over beta
	82	fBeta*=fAlpha;
	83	Annealing(nk);
	84	NumCl(nc,nk);
	85	}while((fBeta<betaStop \|\| nc<4) && nc<fNclustMax);
	86
	87	Int_t *xx=new Int_t[fNin];
	88	EndDetAnn(nk,xx,dEtSum);
	89	StoreJets(nk,xx);
	90	delete [] xx;
	91
	92	}
	93
	94	//-----------------------------------------------------------------------------------
	95	void AliDAJetFinder::InitDetAnn(Double_t &dEtSum)
	96	{
	97	//Initialise the variables used by the algorithm
	98	fBeta=0.1;
	99	TClonesArray *lvArray = fReader->GetMomentumArray();
	100	fNin = lvArray->GetEntries();
9e4cc50d	101	fNclustMax= ((AliDAJetHeader*)fHeader)->GetFixedCl() ?
	102	((AliDAJetHeader*)fHeader)->GetNclustMax()
	103	:
	104	TMath::Max((Int_t)TMath::Sqrt(fNin),5);
7c679be0	105	fXEta=new Double_t[fNin]; fXPhi=new Double_t[fNin];
	106	fPx = new TVectorD(fNin);
	107	for (Int_t iIn=0; iIn<fNin; iIn++){
	108	TLorentzVector lv=(TLorentzVector)lvArray->At(iIn);
	109	fXEta[iIn] = lv->Eta();
	110	fXPhi[iIn] = lv->Phi()<0 ? lv->Phi() + 2*TMath::Pi() : lv->Phi();
	111	(*fPx)(iIn)=lv->Pt();
	112	dEtSum+=(*fPx)(iIn);
	113	}
	114	for (Int_t iIn=0; iIn<fNin; iIn++) (fPx)(iIn)=(fPx)(iIn)/dEtSum;
	115
	116	Int_t njdim=2*fNclustMax+1;
	117	fPyx = new TMatrixD(fNin,njdim);
	118	fY = new TMatrixD(4,njdim);
	119	fPy= new TVectorD(njdim);
	120	fY->Zero();fPyx->Zero();fPy->Zero();
	121	(*fPy)(0)=1;
	122	TMatrixDColumn(*fPyx,0)=1;
	123	Double_t ypos=0,xpos=0;
	124	for (Int_t iIn=0; iIn<fNin; iIn++){
	125	(fY)(0,0)+=(fPx)(iIn)*fXEta[iIn];
	126	ypos+=(fPx)(iIn)TMath::Sin(fXPhi[iIn]);
	127	xpos+=(fPx)(iIn)TMath::Cos(fXPhi[iIn]);
	128	}
	129	(fY)(1,0)=(atan2(ypos,xpos)>0) ? atan2(ypos,xpos) : atan2(ypos,xpos)+2TMath::Pi();
	130	lvArray->Delete();
	131	}
	132
	133	//-----------------------------------------------------------------------------------
	134	void AliDAJetFinder::DoubleClusters(Int_t nc,Int_t &nk)
	135	{
	136	for(Int_t iClust=0; iClust<nc; iClust++){
	137	(fPy)(iClust)=(fPy)(iClust)/2;
	138	(fPy)(nc+iClust)=(fPy)(iClust);
	139	for(Int_t iComp=0; iComp<3; iComp++) (fY)(iComp,nc+iClust)=(fY)(iComp,iClust);
	140	}
	141	nk=2*nc;
	142	}
	143
	144	//-----------------------------------------------------------------------------------
	145	void AliDAJetFinder::Annealing(Int_t nk)
	146	{
	147	// Main part of the algorithm
	148	const Double_t pi=TMath::Pi();
	149	TVectorD *py = new TVectorD(nk);
	150	TVectorD *p = new TVectorD(nk);
	151	TMatrixD *y = new TMatrixD(4,nk);
	152	TMatrixD *y1 = new TMatrixD(4,nk);
	153	TMatrixD *ry = new TMatrixD(2,nk);
	154	Double_t Dist(TVectorD,TVectorD);
	155
9e4cc50d	156	Double_t df[2]={((AliDAJetHeader*)fHeader)->GetEtaCut(),pi};
7c679be0	157	TVectorD vPart(2);
	158	Double_t *m = new Double_t[nk];
	159	Double_t chi,chi1;
	160	do{
	161	Int_t nloop=0;
	162	for (Int_t iClust=0; iClust<nk; iClust++){
	163	for (Int_t i=0; i<3; i++)(y1)(i,iClust)=(fY)(i,iClust);
	164	(py)(iClust)=(fPy)(iClust);
	165	}
	166	//perturbation of codevectors
	167	Double_t seed=1000000*gRandom->Rndm(24);
	168	ry->Randomize(-0.5,0.5,seed);
	169	for (Int_t i=0; i<2; i++){
	170	for (Int_t iClust=0; iClust<nk/2; iClust++)
	171	(y1)(i,iClust)+=((ry)(i,iClust)+TMath::Sign(0.5,(ry)(i,iClust)))fDelta*df[i];
	172	for (Int_t iClust=nk/2; iClust<nk; iClust++)
	173	(y1)(i,iClust)-=((ry)(i,iClust-nk/2)+TMath::Sign(0.5,(ry)(i,iClust-nk/2)))fDelta*df[i];
	174	}
	175	do{
	176	//recalculate conditional probabilities
	177	nloop++;
	178	for (Int_t iIn=0; iIn<fNin; iIn++){
	179	vPart(0)=fXEta[iIn]; vPart(1)=fXPhi[iIn];
	180	for(Int_t iClust=0; iClust<nk; iClust++){
	181	(fPyx)(iIn,iClust)=-log((py)(iClust))+fBetaDist(vPart,TMatrixDColumn(y1,iClust));
	182	m[iClust]=(*fPyx)(iIn,iClust);
	183	}
	184	Double_t pyxNorm=0;
	185	Double_t minPyx=TMath::MinElement(nk,m);
	186	for (Int_t iClust=0; iClust<nk; iClust++){
	187	(*fPyx)(iIn,iClust)-=minPyx;
	188	(fPyx)(iIn,iClust)=exp(-(fPyx)(iIn,iClust));
	189	pyxNorm+=(*fPyx)(iIn,iClust);
	190	}
	191	for (Int_t iClust=0; iClust<nk; iClust++) (*fPyx)(iIn,iClust)/=pyxNorm;
	192	}
	193	p->Zero();
	194	y->Zero();
	195	//recalculate codevectors
	196	for (Int_t iClust=0; iClust<nk; iClust++){
	197	Double_t xpos=0,ypos=0,pxy;
	198	for (Int_t iIn=0; iIn<fNin; iIn++) (p)(iClust)+=(fPx)(iIn)(fPyx)(iIn,iClust);
	199	for (Int_t iIn=0; iIn<fNin; iIn++){
	200	pxy=(fPx)(iIn)(fPyx)(iIn,iClust)/(p)(iClust);
	201	ypos+=pxy*TMath::Sin(fXPhi[iIn]);
	202	xpos+=pxy*TMath::Cos(fXPhi[iIn]);
	203	(y)(0,iClust)+=pxyfXEta[iIn];
	204	}
	205	(y)(1,iClust)=(atan2(ypos,xpos)>0) ? atan2(ypos,xpos) : atan2(ypos,xpos)+2pi;
	206	}
	207	//verify codevectors' stability
	208	chi=0;
	209	for (Int_t iClust=0; iClust<nk; iClust++){
	210	chi1=TMath::CosH((y1)(0,iClust)-(y)(0,iClust))-TMath::Cos((y1)(1,iClust)-(y)(1,iClust));
	211	chi1/=(2TMath::CosH((y1)(0,iClust))TMath::CosH((y)(0,iClust)));
	212	chi1*=chi1;
	213	if (chi1>chi) chi=chi1;
	214	}
	215	chi=TMath::Sqrt(chi);
	216	for (Int_t iClust=0; iClust<nk; iClust++){
	217	for (Int_t i=0; i<2; i++) (y1)(i,iClust)=(y)(i,iClust);
	218	(py)(iClust)=(p)(iClust);
	219	}
	220	if (nloop>fNloopMax){
221	if (chi<fEpsMax \|\| nloop>500) break;
222	}
223	}while (chi>fEps);
224	}while (chi>fEpsMax);
225	for (Int_t iClust=0; iClust<nk; iClust++){ //set codevectors and probability equal to those calculated
226	for (Int_t i=0; i<2; i++) (fY)(i,iClust)=(y)(i,iClust);
227	(fPy)(iClust)=(p)(iClust);
228	}
229	delete py;
230	delete p;
231	delete y;
232	delete y1;
233	delete ry;
234	delete [] m;
235	}
236
237	//-----------------------------------------------------------------------------------
238	void AliDAJetFinder::NumCl(Int_t &nc,Int_t &nk)
239	{
240	static Bool_t growcl=true;
241
242	if (nk==2) growcl=true;
243	if (growcl){
244	//verify if two codevectors are equal within fAvDist
245	Int_t *nSame = new Int_t[nk];
246	Int_t *iSame = new Int_t[nk];
247	Int_t *cont = new Int_t[nk];
248	for (Int_t iClust=0; iClust<nk; iClust++) cont[iClust]=new Int_t[nk],iSame[iClust]=new Int_t[nk];
249	for (Int_t iClust=0; iClust<nk; iClust++){
250	iSame[iClust][iClust]=1;
251	for (Int_t iClust1=iClust+1; iClust1<nk; iClust1++){
252	Double_t eta = (fY)(0,iClust) ; Double_t phi = (fY)(1,iClust);
253	Double_t eta1 = (fY)(0,iClust1); Double_t phi1 = (fY)(1,iClust1);
254	Double_t distCl=(TMath::CosH(eta-eta1)-TMath::Cos(phi-phi1))/(2TMath::CosH(eta)TMath::CosH(eta1));
255	if (distCl < fAvDist) iSame[iClust][iClust1]=iSame[iClust1][iClust]=1;
852db00e	256	else iSame[iClust][iClust1]=iSame[iClust1][iClust]=0;
7c679be0	257	}
	258	}
	259	ReduceClusters(iSame,nk,nc,cont,nSame);
	260	if (nc >= fNclustMax) growcl=false;
	261	//recalculate the nc distinct codevectors
	262	TMatrixD pyx = new TMatrixD(fNin,2nc);
	263	TVectorD *py = new TVectorD(nk);
	264	TMatrixD *y1 = new TMatrixD(3,nk);
	265	for (Int_t iClust=0; iClust<nc; iClust++){
	266	for(Int_t j=0; j<nSame[iClust]; j++){
	267	Int_t iClust1 = cont[iClust][j];
	268	for (Int_t iIn=0; iIn<fNin; iIn++) (pyx)(iIn,iClust)+=(fPyx)(iIn,iClust1);
	269	(py)(iClust)+=(fPy)(iClust1);
	270	for (Int_t i=0; i<2; i++) (y1)(i,iClust)+=(fY)(i,iClust1);
	271	}
	272	for (Int_t i=0; i<2; i++) (*y1)(i,iClust)/=nSame[iClust];
	273	}
852db00e	274	for (Int_t iClust=0; iClust<nk; iClust++) delete [] cont[iClust], delete [] iSame[iClust];
	275	delete [] iSame;
	276	delete [] cont;
	277	delete [] nSame;
7c679be0	278	if (nc > nk/2){
	279	for (Int_t iClust=0; iClust<nc; iClust++){
	280	for (Int_t iIn=0; iIn<fNin; iIn++) (fPyx)(iIn,iClust)=(pyx)(iIn,iClust);
	281	for (Int_t iComp=0; iComp<2; iComp++) (fY)(iComp,iClust)=(y1)(iComp,iClust);
	282	(fPy)(iClust)=(py)(iClust);
	283	}
	284	nk=nc;
	285	if (growcl) DoubleClusters(nc,nk);
	286	}
7c679be0	287	delete pyx;
	288	delete py;
	289	delete y1;
	290	}
	291
	292	}
	293
	294	//-----------------------------------------------------------------------------------
	295	void AliDAJetFinder::ReduceClusters(Int_t iSame,Int_t nc,Int_t &ncout,Int_t cont,Int_t *nSameOut)
	296	{
	297	Int_t *nSame = new Int_t[nc];
	298	Int_t *iperm = new Int_t[nc];
	299	Int_t *go = new Int_t[nc];
	300	for (Int_t iCl=0; iCl<nc; iCl++){
	301	nSame[iCl]=0;
852db00e	302	for (Int_t jCl=0; jCl<nc; jCl++) nSame[iCl]+=iSame[iCl][jCl], cont[iCl][jCl]=0;
7c679be0	303	iperm[iCl]=iCl;
	304	go[iCl]=1;
	305	}
	306	TMath::Sort(nc,nSame,iperm,true);
	307	Int_t l=0;
	308	for (Int_t iCl=0; iCl<nc; iCl++){
	309	Int_t k=iperm[iCl];
	310	if (go[k] == 1){
	311	Int_t m=0;
	312	for (Int_t jCl=0; jCl<nc; jCl++){
	313	if (iSame[k][jCl] == 1){
	314	cont[l][m]=jCl;
	315	go[jCl]=0;
	316	m++;
	317	}
	318	}
	319	nSameOut[l]=m;
	320	l++;
	321	}
	322	}
	323	ncout=l;
	324	delete [] nSame;
	325	delete [] iperm;
	326	delete [] go;
	327	}
	328
	329	//-----------------------------------------------------------------------------------
	330	void AliDAJetFinder::EndDetAnn(Int_t &nk,Int_t *xx,Double_t etx)
	331	{
	332	//now assign each particle to only one cluster
	333	Double_t *clusters=new Double_t[nk];
	334	for (Int_t iIn=0; iIn<fNin; iIn++){
	335	for (Int_t iClust=0; iClust<nk; iClust++) clusters[iClust]=(*fPyx)(iIn,iClust);
	336	xx[iIn]=TMath::LocMax(nk,clusters);
	337	}
	338	delete [] clusters;
	339
	340	//recalculate codevectors, having all p(y\|x)=0 or 1
	341	fY->Zero();
	342	fPyx->Zero();
	343	fPy->Zero();
	344	for (Int_t iIn=0; iIn<fNin; iIn++){
	345	Int_t iClust=xx[iIn];
	346	(*fPyx)(iIn,iClust)=1;
	347	(fPy)(iClust)+=(fPx)(iIn);
	348	(fY)(0,iClust)+=(fPx)(iIn)*fXEta[iIn];
	349	(fY)(3,iClust)+=(fPx)(iIn)*etx;
	350	}
	351	Int_t k=0;
	352	for (Int_t iClust=0; iClust<nk; iClust++){
	353	if ((*fPy)(iClust)>0){
	354	Double_t xpos=0,ypos=0,pxy;
	355	for (Int_t iIn=0; iIn<fNin; iIn++){
	356	pxy=(fPx)(iIn)(fPyx)(iIn,iClust)/(fPy)(iClust);
	357	ypos+=pxy*TMath::Sin(fXPhi[iIn]);
	358	xpos+=pxy*TMath::Cos(fXPhi[iIn]);
	359	if (xx[iIn]==iClust) xx[iIn]=k;
	360	}
	361	(fY)(0,k)=(fY)(0,iClust)/(*fPy)(iClust);
	362	(fY)(1,k)=(atan2(ypos,xpos)>0) ? atan2(ypos,xpos) : atan2(ypos,xpos)+2TMath::Pi();
	363	(fY)(3,k)=(fY)(3,iClust);
	364	k++;
	365	}
	366	}
367	nk=k;
368	}
369
370	//-----------------------------------------------------------------------------------
371	void AliDAJetFinder::StoreJets(Int_t nk,Int_t *xx)
372	{
373	//evaluate significant clusters properties
374	const Double_t pi=TMath::Pi();
9e4cc50d	375	Double_t dMeanDist=TMath::Sqrt(4((AliDAJetHeader)fHeader)->GetEtaCut()*pi/fNin);
7c679be0	376	Bool_t *isJet = new Bool_t[nk];
	377	Double_t *etNoBg= new Double_t[nk];
	378	Double_t *dDeltaEta=new Double_t[nk];
	379	Double_t *dDeltaPhi=new Double_t[nk];
	380	Double_t *surf = new Double_t[nk];
	381	Double_t *etDens= new Double_t[nk];
	382	for (Int_t iClust=0; iClust<nk; iClust++){ //clusters loop
	383	isJet[iClust]=false;
	384	Double_t dEtaMin=10.,dEtaMax=-10.,dPhiMin=10.,dPhiMax=0.;
	385	for (Int_t iIn=0; iIn<fNin; iIn++){
	386	if (xx[iIn]!=iClust) continue;
	387	if (fXEta[iIn] < dEtaMin) dEtaMin=fXEta[iIn];
	388	if (fXEta[iIn] > dEtaMax) dEtaMax=fXEta[iIn];
	389	Double_t dPhi=fXPhi[iIn]-(*fY)(1,iClust);
	390	if (dPhi > pi ) dPhi-=2*pi;
	391	else if (dPhi < (-1)pi) dPhi+=2pi;
	392	if (dPhi < dPhiMin) dPhiMin=dPhi;
	393	else if (dPhi > dPhiMax) dPhiMax=dPhi;
	394	}
	395	dDeltaEta[iClust]=dEtaMax-dEtaMin+dMeanDist;
	396	dDeltaPhi[iClust]=dPhiMax-dPhiMin+dMeanDist;
	397	surf[iClust]=0.25pidDeltaEta[iClust]*dDeltaPhi[iClust];
	398	etDens[iClust]=(*fY)(3,iClust)/surf[iClust];
	399	}
	400
9e4cc50d	401	if (((AliDAJetHeader*)fHeader)->GetSelJets()){
7c679be0	402	for (Int_t iClust=0; iClust<nk; iClust++){
	403	if (!isJet[iClust]){
	404	Double_t etDensMed=0.;
	405	Double_t etDensSqr=0.;
	406	Int_t norm=0;
	407	for (Int_t iClust1=0; iClust1<nk; iClust1++){
	408	if(iClust1!=iClust){
	409	etDensMed+=etDens[iClust1];
	410	etDensSqr+=TMath::Power(etDens[iClust1],2);
	411	norm++;
	412	}
	413	}
	414	etDensMed/=TMath::Max(norm,1);
	415	etDensSqr/=TMath::Max(norm,1);
	416	Double_t deltaEtDens=TMath::Sqrt(etDensSqr-TMath::Power(etDensMed,2));
	417	if ((fY)(3,iClust) > (etDensMed+deltaEtDens)surf[iClust]) isJet[iClust]=kTRUE;
	418	etNoBg[iClust]=(fY)(3,iClust)-etDensMedsurf[iClust];
	419	}
	420	}
	421	for (Int_t iClust=0; iClust<nk; iClust++){
	422	if (isJet[iClust]){
	423	Double_t etDensMed=0;
9e4cc50d	424	Double_t extSurf=4((AliDAJetHeader)fHeader)->GetEtaCut()*pi;
7c679be0	425	for (Int_t iClust1=0; iClust1<nk; iClust1++){
	426	if (!isJet[iClust1]) etDensMed+=(*fY)(3,iClust1);
	427	else extSurf-=surf[iClust1];
	428	}
	429	etDensMed/=extSurf;
	430	etNoBg[iClust]=(fY)(3,iClust)-etDensMedsurf[iClust];
10bd125d	431	if (etNoBg[iClust]<((AliDAJetHeader*)fHeader)->GetEtMin()){
	432	isJet[iClust]=kFALSE;
	433	iClust=-1;
	434	}
7c679be0	435	}
	436	}
	437	} else {
	438	for (Int_t iClust=0; iClust<nk; iClust++) isJet[iClust]=true;
	439	}
	440	delete [] etDens;
	441	delete [] surf;
	442
	443	//now add selected jets to the list
	444	Int_t *inJet = new Int_t[fNin];
	445	TRefArray *refs = 0;
	446	Bool_t fromAod = !strcmp(fReader->ClassName(),"AliJetAODReader");
	447	if (fromAod) refs = fReader->GetReferences();
	448	for (Int_t iClust=0; iClust<nk; iClust++){ //clusters loop
	449	if (isJet[iClust]){ //choose cluster
	450	Float_t px,py,pz,en;
	451	px = (fY)(3,iClust)TMath::Cos((*fY)(1,iClust));
	452	py = (fY)(3,iClust)TMath::Sin((*fY)(1,iClust));
	453	pz = (fY)(3,iClust)/TMath::Tan(2.0 TMath::ATan(TMath::Exp(-(*fY)(0,iClust))));
	454	en = TMath::Sqrt(px * px + py * py + pz * pz);
	455	AliAODJet jet(px, py, pz, en);
	456	if (fromAod)
	457	for (Int_t iIn=0; iIn<fNin; iIn++) if (xx[iIn]==iClust) jet.AddTrack(refs->At(iIn));
	458	AddJet(jet);
	459	printf("jet %d, Eta: %f, Phi: %f, Et: %f\n",iClust,jet.Eta(),jet.Phi(),jet.Pt());
	460	}
	461	}
	462	delete [] dDeltaEta; delete [] dDeltaPhi;
	463	delete [] etNoBg;
	464	delete [] isJet;
	465	delete [] inJet;
	466	}
	467
	468	//-----------------------------------------------------------------------------------
	469	Double_t Dist(TVectorD x,TVectorD y)
	470	{
	471	// Squared distance
	472	const Double_t pi=TMath::Pi();
	473	Double_t dphi=TMath::Abs(x(1)-y(1));
	474	if (dphi > pi) dphi=2*pi-dphi;
	475	Double_t dist=pow(x(0)-y(0),2)+pow(dphi,2);
	476	return dist;
	477	}