NewFitterGSL Class Reference

A kinematic fitter using the Newton-Raphson method to solve the equations. More...

#include <NewFitterGSL.h>

Inheritance diagram for NewFitterGSL:

Collaboration diagram for NewFitterGSL:

Public Types
enum	{   NPARMAX = 50 ,   NCONMAX = 10 ,   NUNMMAX = 10 }
enum	{ NITMAX = 100 }

Public Member Functions
	NewFitterGSL ()
	Constructor.
virtual	~NewFitterGSL ()
	Virtual destructor.
virtual double	fit () override
	The fit method, returns the fit probability.
virtual int	getError () const override
	Get the error code of the last fit: 0=OK, 1=failed.
virtual double	getProbability () const override
	Get the fit probability of the last fit.
virtual double	getChi2 () const override
	Get the chi**2 of the last fit.
virtual int	getDoF () const override
	Get the number of degrees of freedom of the last fit.
virtual int	getIterations () const override
	Get the number of iterations of the last fit.
virtual int	getNcon () const
	Get the number of hard constraints of the last fit.
virtual int	getNsoft () const
	Get the number of soft constraints of the last fit.
virtual int	getNpar () const
	Get the number of all parameters of the last fit.
virtual int	getNunm () const
	Get the number of unmeasured parameters of the last fit.
virtual bool	initialize () override
	Initialize the fitter.
virtual void	setDebug (int debuglevel)
	Set the Debug Level.
virtual int	determineLambdas (gsl_vector *vecxnew, const gsl_matrix *MatM, const gsl_vector *vecx, gsl_matrix *MatW, gsl_vector *vecw, double eps=0)
	Determine best lambda values.
virtual void	calc2ndOrderCorr (gsl_vector *vecdxhat, const gsl_vector *vecxnew, const gsl_matrix *MatM, gsl_matrix *MatW, gsl_vector *vecw, double eps=0)
	Calculate 2nd order correction step.
virtual gsl_matrix_view	calcZ (int &rankA, gsl_matrix *MatW1, gsl_matrix *MatW2, gsl_vector *vecw1, gsl_vector *vecw2, gsl_permutation *permW, double eps=0)
	Calculate null space of constraints; return value is a matrix view of MatW1, with column vectors spanning null(A)
virtual gsl_matrix_view	calcReducedHessian (int &rankH, gsl_matrix *MatW1, const gsl_vector *vecx, gsl_matrix *MatW2, gsl_matrix *MatW3, gsl_vector *vecw1, gsl_vector *vecw2, gsl_permutation *permW, double eps=0)
	Calculate reduced Hessian; return value is a matrix view of MatW1 giving the reduced Hessian.
virtual gsl_vector_view	calcReducedHessianEigenvalues (int &rankH, gsl_matrix *MatW1, const gsl_vector *vecx, gsl_matrix *MatW2, gsl_matrix *MatW3, gsl_vector *vecw1, gsl_vector *vecw2, gsl_permutation *permW, gsl_eigen_symm_workspace *eigenws, double eps=0)
virtual double	calcChi2 ()
	Calculate the chi2.
void	fillx (gsl_vector *vecx)
void	fillperr (gsl_vector *vece)
void	assembleM (gsl_matrix *MatM, const gsl_vector *vecx, bool errorpropagation=false)
void	assembleG (gsl_matrix *MatM, const gsl_vector *vecx)
void	scaleM (gsl_matrix *MatMscal, const gsl_matrix *MatM, const gsl_vector *vece)
void	assembley (gsl_vector *vecy, const gsl_vector *vecx)
void	scaley (gsl_vector *vecyscal, const gsl_vector *vecy, const gsl_vector *vece)
int	assembleChi2Der (gsl_vector *vecy)
void	addConstraints (gsl_vector *vecy)
void	assembleConstDer (gsl_matrix *MatM)
int	calcNewtonDx (gsl_vector *vecdx, gsl_vector *vecdxscal, gsl_vector *vecx, const gsl_vector *vece, gsl_matrix *MatM, gsl_matrix *MatMscal, gsl_vector *vecy, gsl_vector *vecyscal, gsl_matrix *MatW, gsl_matrix *MatW2, gsl_permutation *permW, gsl_vector *vecw)
int	calcLimitedDx (double &alpha, double &mu, gsl_vector *vecxnew, int imode, gsl_vector *vecx, gsl_vector *vecdxhat, const gsl_vector *vecdx, const gsl_vector *vecdxscal, const gsl_vector *vece, const gsl_matrix *MatM, const gsl_matrix *MatMscal, gsl_matrix *MatW, gsl_vector *vecw)
int	doLineSearch (double &alpha, gsl_vector *vecxnew, int imode, double phi0, double dphi0, double eta, double zeta, double mu, const gsl_vector *vecx, const gsl_vector *vecdx, const gsl_vector *vece, gsl_vector *vecw)
double	calcMu (const gsl_vector *vecx, const gsl_vector *vece, const gsl_vector *vecdx, const gsl_vector *vecdxscal, const gsl_vector *xnew, const gsl_matrix *MatM, const gsl_matrix *MatMscal, gsl_vector *vecw)
double	meritFunction (double mu, const gsl_vector *vecx, const gsl_vector *vece)
double	meritFunctionDeriv (double mu, const gsl_vector *vecx, const gsl_vector *vece, const gsl_vector *vecdx, gsl_vector *vecw)
bool	updateParams (gsl_vector *vecx)
int	invertM ()
int	calcCovMatrix (gsl_matrix *MatW, gsl_permutation *permW, gsl_vector *vecx)
double	calcpTLp (const gsl_vector *vecdx, const gsl_matrix *MatM, gsl_vector *vecw)
int	solveSystem (gsl_vector *vecdxscal, double &detW, const gsl_vector *vecyscal, const gsl_matrix *MatMscal, gsl_matrix *MatW, gsl_matrix *MatW2, gsl_vector *vecw, double epsLU, double epsSV)
	solve system of equations Mscal*dxscal = yscal
int	solveSystemLU (gsl_vector *vecdxscal, double &detW, const gsl_vector *vecyscal, const gsl_matrix *MatMscal, gsl_matrix *MatW, gsl_vector *vecw, double eps)
	solve system of equations Mscal*dxscal = yscal using LU decomposition
int	solveSystemSVD (gsl_vector *vecdxscal, const gsl_vector *vecyscal, const gsl_matrix *MatMscal, gsl_matrix *MatW, gsl_matrix *MatW2, gsl_vector *vecw, double eps)
	solve system of equations Mscal*dxscal = yscal using SVD decomposition
virtual void	addFitObject (BaseFitObject *fitobject_)
virtual void	addFitObject (BaseFitObject &fitobject_)
virtual void	addConstraint (BaseConstraint *constraint_)
virtual void	addConstraint (BaseConstraint &constraint_)
virtual void	addHardConstraint (BaseHardConstraint *constraint_)
virtual void	addHardConstraint (BaseHardConstraint &constraint_)
virtual void	addSoftConstraint (BaseSoftConstraint *constraint_)
virtual void	addSoftConstraint (BaseSoftConstraint &constraint_)
virtual std::vector< BaseFitObject * > *	getFitObjects ()
virtual std::vector< BaseHardConstraint * > *	getConstraints ()
virtual std::vector< BaseSoftConstraint * > *	getSoftConstraints ()
virtual void	reset ()
virtual BaseTracer *	getTracer ()
virtual const BaseTracer *	getTracer () const
virtual void	setTracer (BaseTracer *newTracer)
virtual void	setTracer (BaseTracer &newTracer)
virtual const double *	getGlobalCovarianceMatrix (int &idim) const
virtual double *	getGlobalCovarianceMatrix (int &idim)

Static Public Member Functions
static void	ini_gsl_permutation (gsl_permutation *&p, unsigned int size)
static void	ini_gsl_vector (gsl_vector *&v, int unsigned size)
static void	ini_gsl_matrix (gsl_matrix *&m, int unsigned size1, unsigned int size2)
static void	debug_print (const gsl_matrix *m, const char *name)
static void	debug_print (const gsl_vector *v, const char *name)
static void	add (gsl_vector *vecz, const gsl_vector *vecx, double a, const gsl_vector *vecy)
static bool	isfinite (const gsl_vector *vec)
static bool	isfinite (const gsl_matrix *mat)
static void	MoorePenroseInverse (gsl_matrix *Ainv, gsl_matrix *A, gsl_matrix *W, gsl_vector *w, double eps=0)
	Compute the Moore-Penrose pseudo-inverse A+ of A, using SVD.

Public Attributes
int	npar
	total number of parameters
int	ncon
	total number of hard constraints
int	nsoft
	total number of soft constraints
int	nunm
	total number of unmeasured parameters
int	ierr
	Error status.
int	nit
	Number of iterations.
double	fitprob
	fit probability
double	chi2
	final chi2
unsigned int	idim
gsl_vector *	x
gsl_vector *	xold
gsl_vector *	xnew
gsl_vector *	dx
gsl_vector *	dxscal
gsl_vector *	y
gsl_vector *	yscal
gsl_vector *	perr
gsl_vector *	v1
gsl_vector *	v2
gsl_matrix *	M
gsl_matrix *	Mscal
gsl_matrix *	W
gsl_matrix *	W2
gsl_matrix *	W3
gsl_matrix *	M1
gsl_matrix *	M2
gsl_matrix *	M3
gsl_matrix *	M4
gsl_matrix *	M5
gsl_matrix *	CC
gsl_matrix *	CC1
gsl_matrix *	CCinv
gsl_permutation *	permW
gsl_eigen_symm_workspace *	eigenws
unsigned int	eigenwsdim
double	chi2best
double	chi2new
double	chi2old
double	fvalbest
double	scale
double	scalebest
double	stepsize
double	stepbest
double	scalevals [NITMAX]
double	fvals [NITMAX]
int	imerit
bool	try2ndOrderCorr
int	debug
std::map< std::string, double >	traceValues

Protected Types
typedef std::vector< BaseFitObject * >	FitObjectContainer
typedef std::vector< BaseHardConstraint * >	ConstraintContainer
typedef std::vector< BaseSoftConstraint * >	SoftConstraintContainer
typedef FitObjectContainer::iterator	FitObjectIterator
typedef ConstraintContainer::iterator	ConstraintIterator
typedef SoftConstraintContainer::iterator	SoftConstraintIterator

Protected Attributes
FitObjectContainer	fitobjects
ConstraintContainer	constraints
SoftConstraintContainer	softconstraints
int	covDim
	dimension of global covariance matrix
double *	cov
	global covariance matrix of last fit problem
bool	covValid
	Flag whether global covariance is valid.
BaseTracer *	tracer

Detailed Description

A kinematic fitter using the Newton-Raphson method to solve the equations.

This class implements a kinematic fitter using the Newton-Raphson method to solve the system of equations arising from the Lagrange multiplier method

Author: Benno List Last update:

Date

2011/05/03 13:16:41

by:

Author

blist

Changelog:

• 15.11.2010 First version

Definition at line 51 of file NewFitterGSL.h.

Member Typedef Documentation

ConstraintContainer

typedef std::vector<BaseHardConstraint*> ConstraintContainer

protectedinherited

Definition at line 92 of file BaseFitter.h.

ConstraintIterator

typedef ConstraintContainer::iterator ConstraintIterator

protectedinherited

Definition at line 96 of file BaseFitter.h.

FitObjectContainer

typedef std::vector<BaseFitObject*> FitObjectContainer

protectedinherited

Definition at line 91 of file BaseFitter.h.

FitObjectIterator

typedef FitObjectContainer::iterator FitObjectIterator

protectedinherited

Definition at line 95 of file BaseFitter.h.

SoftConstraintContainer

typedef std::vector<BaseSoftConstraint*> SoftConstraintContainer

protectedinherited

Definition at line 93 of file BaseFitter.h.

SoftConstraintIterator

typedef SoftConstraintContainer::iterator SoftConstraintIterator

protectedinherited

Definition at line 97 of file BaseFitter.h.

Member Enumeration Documentation

anonymous enum

anonymous enum

Definition at line 252 of file NewFitterGSL.h.

{NPARMAX = 50, NCONMAX = 10, NUNMMAX = 10};

anonymous enum

anonymous enum

Definition at line 370 of file NewFitterGSL.h.

{NITMAX = 100};

Constructor & Destructor Documentation

NewFitterGSL()

NewFitterGSL

(

)

Constructor.

Definition at line 55 of file NewFitterGSL.cc.

      : npar(0), ncon(0), nsoft(0), nunm(0), ierr(0), nit(0),
        fitprob(0), chi2(0), idim(0), x(nullptr), xold(nullptr), xnew(nullptr),
        // xbest(0),
        dx(nullptr), dxscal(nullptr),
        //grad(0),
        y(nullptr), yscal(nullptr),
        perr(nullptr), v1(nullptr), v2(nullptr),
        //Meval (0),
        M(nullptr), Mscal(nullptr), W(nullptr), W2(nullptr), W3(nullptr),
        M1(nullptr), M2(nullptr), M3(nullptr), M4(nullptr), M5(nullptr),
        //Mevec (0),
        CC(nullptr), CC1(nullptr), CCinv(nullptr),
        permW(nullptr), eigenws(nullptr), eigenwsdim(0),
        chi2best(0), chi2new(0), chi2old(0), fvalbest(0),
        scale(0), scalebest(0), stepsize(0), stepbest(0),
        scalevals{0}, fvals{0},
        imerit(1),
        try2ndOrderCorr(true),
        debug(debuglevel)
    {}

~NewFitterGSL()

~NewFitterGSL ( )

virtual

Virtual destructor.

Definition at line 78 of file NewFitterGSL.cc.

    {
 
      if (x) gsl_vector_free(x);
      if (xold) gsl_vector_free(xold);
      if (xnew) gsl_vector_free(xnew);
//   if (xbest) gsl_vector_free (xbest);
      if (dx) gsl_vector_free(dx);
      if (dxscal) gsl_vector_free(dxscal);
//   if (grad) gsl_vector_free (grad);
      if (y) gsl_vector_free(y);
      if (yscal) gsl_vector_free(yscal);
      if (perr) gsl_vector_free(perr);
      if (v1) gsl_vector_free(v1);
      if (v2) gsl_vector_free(v2);
//   if (Meval) gsl_vector_free (Meval);
      if (M) gsl_matrix_free(M);
      if (Mscal) gsl_matrix_free(Mscal);
      if (W) gsl_matrix_free(W);
      if (W2) gsl_matrix_free(W2);
      if (W3) gsl_matrix_free(W3);
      if (M1) gsl_matrix_free(M1);
      if (M2) gsl_matrix_free(M2);
      if (M3) gsl_matrix_free(M3);
      if (M4) gsl_matrix_free(M4);
      if (M5) gsl_matrix_free(M5);
//   if (Mevec) gsl_matrix_free (Mevec);
      if (CC) gsl_matrix_free(CC);
      if (CC1) gsl_matrix_free(CC1);
      if (CCinv) gsl_matrix_free(CCinv);
      if (permW) gsl_permutation_free(permW);
      if (eigenws) gsl_eigen_symm_free(eigenws);
      x = nullptr;
      xold = nullptr;
      xnew = nullptr;
//     xbest=0;
      dx = nullptr;
      dxscal = nullptr;
//     grad=0;
      y = nullptr;
      yscal = nullptr;
      perr = nullptr;
      v1 = nullptr;
      v2 = nullptr;
//     Meval=0;
      M = nullptr;
      Mscal = nullptr;
      W = nullptr;
      W2 = nullptr;
      W3 = nullptr;
      M1 = nullptr;
      M2 = nullptr;
      M3 = nullptr;
      M4 = nullptr;
      M5 = nullptr;
//   Mevec=0;
      CC = nullptr;
      CC1 = nullptr;
      CCinv = nullptr;
      permW = nullptr;
      eigenws = nullptr; eigenwsdim = 0;
    }

Member Function Documentation

add()

void add ( gsl_vector *  vecz, const gsl_vector *  vecx, double  a, const gsl_vector *  vecy  )

static

Definition at line 485 of file NewFitterGSL.cc.

    {
      assert(vecx);
      assert(vecy);
      assert(vecz);
      assert(vecx->size == vecy->size);
      assert(vecx->size == vecz->size);
 
      gsl_blas_dcopy(vecx, vecz);
      gsl_blas_daxpy(a, vecy, vecz);
    }

addConstraint() [1/2]

void addConstraint ( BaseConstraint &  constraint_ )

virtualinherited

Definition at line 74 of file BaseFitter.cc.

    {
      covValid = false;
      if (auto* hc = dynamic_cast<BaseHardConstraint*>(&constraint_))
        constraints.push_back(hc);
      else if (auto* sc = dynamic_cast<BaseSoftConstraint*>(&constraint_))
        softconstraints.push_back(sc);
    }

addConstraint() [2/2]

void addConstraint ( BaseConstraint *  constraint_ )

virtualinherited

Definition at line 60 of file BaseFitter.cc.

    {
      covValid = false;
 
      if (auto* hc = dynamic_cast<BaseHardConstraint*>(constraint_))
        constraints.push_back(hc);
      else if (auto* sc = dynamic_cast<BaseSoftConstraint*>(constraint_))
        softconstraints.push_back(sc);
      else {
        // illegal constraint
        assert(0);
      }
    }

addConstraints()

void addConstraints

(

gsl_vector *

vecy

)

Definition at line 773 of file NewFitterGSL.cc.

    {
      assert(vecy);
      assert(vecy->size == idim);
 
      // Now add the derivatives w.r.t. to the constraints, i.e. the constraints themselves
      for (auto c : constraints) {
        assert(c);
        int kglobal = c->getGlobalNum();
        gsl_vector_set(vecy, kglobal, c->getValue());
      }
    }

addFitObject() [1/2]

void addFitObject ( BaseFitObject &  fitobject_ )

virtualinherited

Definition at line 54 of file BaseFitter.cc.

    {
      covValid = false;
      fitobjects.push_back(&fitobject_);
    }

addFitObject() [2/2]

void addFitObject ( BaseFitObject *  fitobject_ )

virtualinherited

Definition at line 48 of file BaseFitter.cc.

    {
      covValid = false;
      fitobjects.push_back(fitobject_);
    }

addHardConstraint() [1/2]

void addHardConstraint ( BaseHardConstraint &  constraint_ )

virtualinherited

Definition at line 89 of file BaseFitter.cc.

    {
      covValid = false;
      constraints.push_back(&constraint_);
    }

addHardConstraint() [2/2]

void addHardConstraint ( BaseHardConstraint *  constraint_ )

virtualinherited

Definition at line 83 of file BaseFitter.cc.

    {
      covValid = false;
      constraints.push_back(constraint_);
    }

addSoftConstraint() [1/2]

void addSoftConstraint ( BaseSoftConstraint &  constraint_ )

virtualinherited

Definition at line 101 of file BaseFitter.cc.

    {
      covValid = false;
      softconstraints.push_back(&constraint_);
    }

addSoftConstraint() [2/2]

void addSoftConstraint ( BaseSoftConstraint *  constraint_ )

virtualinherited

Definition at line 95 of file BaseFitter.cc.

    {
      covValid = false;
      softconstraints.push_back(constraint_);
    }

assembleChi2Der()

int assembleChi2Der

(

gsl_vector *

vecy

)

Definition at line 750 of file NewFitterGSL.cc.

    {
      assert(vecy);
      assert(vecy->size == idim);
 
      gsl_vector_set_zero(vecy);
      // First, for the parameters
      for (auto fo : fitobjects) {
        assert(fo);
        //  B2INFO("In New assembleChi2Der FitObject:  "<< fo->getName());
        int ifail = fo->addToGlobalChi2DerVector(vecy->block->data, vecy->size);
        if (ifail) return ifail;
      }
 
      // Treat the soft constraints
 
      for (auto bsc : softconstraints) {
        assert(bsc);
        bsc->addToGlobalChi2DerVector(vecy->block->data, vecy->size);
      }
      return 0;
    }

assembleConstDer()

void assembleConstDer

(

gsl_matrix *

MatM

)

Definition at line 786 of file NewFitterGSL.cc.

    {
      assert(MatM);
      assert(MatM->size1 == idim && MatM->size2 == idim);
 
      gsl_matrix_set_zero(MatM);
 
      // The first derivatives of the constraints,
      for (auto c : constraints) {
        assert(c);
        int kglobal = c->getGlobalNum();
        assert(kglobal >= 0 && kglobal < (int)idim);
        c->add1stDerivativesToMatrix(MatM->block->data, MatM->tda);
      }
    }

assembleG()

void assembleG	(	gsl_matrix *	MatM,
		const gsl_vector *	vecx
	)

Definition at line 657 of file NewFitterGSL.cc.

    {
      assert(MatM);
      assert(MatM->size1 == idim && MatM->size2 == idim);
      assert(vecx);
      assert(vecx->size == idim);
 
      gsl_matrix_set_zero(MatM);
      // First, all terms d^2 chi^2/dx1 dx2
      for (auto fo : fitobjects) {
        assert(fo);
        fo->addToGlobalChi2DerMatrix(MatM->block->data, MatM->tda);
      }
 
      // Second,  the second derivatives times the lambda values
      for (auto c : constraints) {
        assert(c);
        int kglobal = c->getGlobalNum();
        assert(kglobal >= 0 && kglobal < (int)idim);
        c->add2ndDerivativesToMatrix(MatM->block->data, MatM->tda, gsl_vector_get(vecx, kglobal));
      }
 
      // Finally, treat the soft constraints
 
      for (auto bsc : softconstraints) {
        assert(bsc);
        bsc->add2ndDerivativesToMatrix(MatM->block->data, MatM->tda);
      }
    }

assembleM()

void assembleM	(	gsl_matrix *	MatM,
		const gsl_vector *	vecx,
		bool	errorpropagation = false
	)

Definition at line 581 of file NewFitterGSL.cc.

    {
      assert(MatM);
      assert(MatM->size1 == idim && MatM->size2 == idim);
      assert(vecx);
      assert(vecx->size == idim);
 
      gsl_matrix_set_zero(MatM);
 
      // First, all terms d^2 chi^2/dx1 dx2
      for (auto fo : fitobjects) {
        assert(fo);
        fo->addToGlobalChi2DerMatrix(MatM->block->data, MatM->tda);
        if (!isfinite(MatM)) {
          B2DEBUG(10, "NewFitterGSL::assembleM: illegal elements in MatM after adding fo " << *fo << ":\n");
          if (debug > 15) debug_print(MatM, "M");
        }
      }
      if (debug > 13) {
        B2INFO("After adding covariances from fit objects:\n");
        //printMy ((double*) M, (double*) y, (int) idim);
        debug_print(MatM, "MatM");
      }
 
      // Second, all terms d^2 chi^2/dlambda dx,
      // i.e. the first derivatives of the constraints,
      // plus the second derivatives times the lambda values
      for (auto c : constraints) {
        assert(c);
        int kglobal = c->getGlobalNum();
        assert(kglobal >= 0 && kglobal < (int)idim);
        c->add1stDerivativesToMatrix(MatM->block->data, MatM->tda);
        if (!isfinite(MatM)) {
          B2DEBUG(10, "NewFitterGSL::assembleM: illegal elements in MatM after adding 1st derivatives of constraint " << *c << ":\n");
          if (debug > 13) debug_print(MatM, "M");
        }
        if (debug > 13) {
          B2INFO("After adding first derivatives of constraint " << c->getName());
          //printMy ((double*) M, (double*) y, (int) idim);
          debug_print(MatM, "MatM");
          B2INFO("errorpropagation = " << errorpropagation);
        }
        // for error propagation after fit,
        //2nd derivatives of constraints times lambda should _not_ be included!
        if (!errorpropagation) c->add2ndDerivativesToMatrix(MatM->block->data, MatM->tda, gsl_vector_get(vecx, kglobal));
        if (!isfinite(MatM)) {
          B2DEBUG(10, "NewFitterGSL::assembleM: illegal elements in MatM after adding 2nd derivatives of constraint " << *c << ":\n");
          if (debug > 13) debug_print(MatM, "MatM");
        }
      }
      if (debug > 13) {
        B2INFO("After adding derivatives of constraints::\n");
        //printMy ((double*) M, (double*) y, (int) idim);
        debug_print(M, "M");
        B2INFO("===========================================::\n");
      }
 
      // Finally, treat the soft constraints
 
      for (auto bsc : softconstraints) {
        assert(bsc);
        bsc->add2ndDerivativesToMatrix(MatM->block->data, MatM->tda);
        if (!isfinite(MatM)) {
          B2DEBUG(10, "NewFitterGSL::assembleM: illegal elements in MatM after adding soft constraint " << *bsc << ":\n");
          if (debug > 13) debug_print(MatM, "M");
        }
      }
      if (debug > 13) {
        B2INFO("After adding soft constraints::\n");
        //printMy ((double*) M, (double*) y, (int) idim);
        debug_print(M, "M");
        B2INFO("===========================================::\n");
      }
 
    }

assembley()

void assembley	(	gsl_vector *	vecy,
		const gsl_vector *	vecx
	)

Definition at line 705 of file NewFitterGSL.cc.

    {
      assert(vecy);
      assert(vecy->size == idim);
      assert(vecx);
      assert(vecx->size == idim);
      gsl_vector_set_zero(vecy);
      // First, for the parameters
      for (auto fo : fitobjects) {
        assert(fo);
//  B2INFO("In New assembley FitObject:  "<< fo->getName());
        fo->addToGlobalChi2DerVector(vecy->block->data, vecy->size);
      }
 
      // Now add lambda*derivatives of constraints,
      // And finally, the derivatives w.r.t. to the constraints, i.e. the constraints themselves
      for (auto c : constraints) {
        assert(c);
        int kglobal = c->getGlobalNum();
        assert(kglobal >= 0 && kglobal < (int)idim);
        c->addToGlobalChi2DerVector(vecy->block->data, vecy->size, gsl_vector_get(vecx, kglobal));
        gsl_vector_set(vecy, kglobal, c->getValue());
      }
 
      // Finally, treat the soft constraints
 
      for (auto bsc : softconstraints) {
        assert(bsc);
        bsc->addToGlobalChi2DerVector(vecy->block->data, vecy->size);
      }
    }

calc2ndOrderCorr()

void calc2ndOrderCorr ( gsl_vector *  vecdxhat, const gsl_vector *  vecxnew, const gsl_matrix *  MatM, gsl_matrix *  MatW, gsl_vector *  vecw, double  eps = 0  )

virtual

Calculate 2nd order correction step.

Parameters

vecdxhat	the correction step
vecxnew	the current state vector (parameters must be set to vecxnew!)
MatM	The matrix of the last Newton step
MatW	work matrix
vecw	work vector
eps	Singular values < eps*(max(abs(s_i))) are set to 0

Definition at line 1633 of file NewFitterGSL.cc.

    {
      assert(vecdxhat);
      assert(vecdxhat->size == idim);
      assert(vecxnew);
      assert(vecxnew->size == idim);
      assert(MatM);
      assert(MatM->size1 == idim && MatM->size2 == idim);
      assert(MatW);
      assert(MatW->size1 == idim && MatW->size2 == idim);
      assert(vecw);
      assert(vecw->size == idim);
      assert(idim == static_cast<unsigned int>(npar + ncon));
 
 
      // Calculate 2nd order correction
      // see Nocedal&Wright (15.36)
      gsl_matrix_const_view AT(gsl_matrix_const_submatrix(MatM, 0,    npar, npar, ncon));
      gsl_matrix_view AAT(gsl_matrix_submatrix(MatW, npar, npar, ncon, ncon));
      gsl_matrix_view ATcopy(gsl_matrix_submatrix(MatW, 0,    npar, npar, ncon));
      gsl_matrix_view& V = AAT;
 
      gsl_vector_set_zero(vecdxhat);
      addConstraints(vecdxhat);
      gsl_vector_view c(gsl_vector_subvector(vecdxhat, npar, ncon));
      gsl_vector_view phat = (gsl_vector_subvector(vecdxhat, 0, npar));
 
      gsl_vector_set_zero(vecw);
      gsl_vector_view AATinvc(gsl_vector_subvector(vecw, npar, ncon));
      gsl_vector_view& s = AATinvc;
 
 
      // AAT = 1*A*A^T + 0*AAT
      gsl_blas_dgemm(CblasTrans, CblasNoTrans, 1, &AT.matrix, &AT.matrix, 0, &AAT.matrix);
 
      // solve AAT * AATinvc = c using the Cholsky factorization method
      gsl_error_handler_t* old_handler =  gsl_set_error_handler_off();
      int cholesky_result = gsl_linalg_cholesky_decomp(&AAT.matrix);
      gsl_set_error_handler(old_handler);
      if (cholesky_result) {
        B2INFO("NewFitterGSL::calc2ndOrderCorr: resorting to SVD");
        // AAT is not positive definite, i.e. A does not have full column rank
        // => use the SVD of AT to solve A lambdanew = gradf
 
        // ATcopy = AT
        gsl_matrix_memcpy(&ATcopy.matrix, &AT.matrix);
 
        // SVD decomposition of Acopy
        gsl_linalg_SV_decomp_jacobi(&ATcopy.matrix, &V.matrix, &s.vector);
        // set small values to zero
        double mins = eps * std::fabs(gsl_vector_get(&s.vector, 0));
        for (int i = 0; i < ncon; ++i) {
          if (std::fabs(gsl_vector_get(&s.vector, i)) <= mins)
            gsl_vector_set(&s.vector, i, 0);
        }
        gsl_linalg_SV_solve(&ATcopy.matrix, &V.matrix, &s.vector, &c.vector, &AATinvc.vector);
      } else {
        gsl_linalg_cholesky_solve(&AAT.matrix, &c.vector, &AATinvc.vector);
      }
 
      // phat = -1*A^T*AATinvc+ 0*phat
      gsl_blas_dgemv(CblasNoTrans, -1, &AT.matrix, &AATinvc.vector, 0, &phat.vector);
      gsl_vector_set_zero(&c.vector);
 
    }

calcChi2()

double calcChi2 ( )

virtual

Calculate the chi2.

Definition at line 416 of file NewFitterGSL.cc.

    {
      chi2 = 0;
      for (auto fo : fitobjects) {
        assert(fo);
        chi2 += fo->getChi2();
      }
      for (auto bsc : softconstraints) {
        assert(bsc);
        chi2 += bsc->getChi2();
      }
      return chi2;
    }

calcCovMatrix()

int calcCovMatrix	(	gsl_matrix *	MatW,
		gsl_permutation *	permW,
		gsl_vector *	vecx
	)

Definition at line 1368 of file NewFitterGSL.cc.

    {
      // Set up equation system M*dadeta + dydeta = 0
      // here, dadeta is d a / d eta, i.e. the derivatives of the fitted
      // parameters a w.r.t. to the measured parameters eta,
      // and dydeta is the derivative of the set of equations
      // w.r.t eta, i.e. simply d^2 chi^2 / d a d eta.
      // Now, if chi2 = (a-eta)^T*Vinv((a-eta), we have simply
      // d^2 chi^2 / d a d eta = - d^2 chi^2 / d a d a
      // and can use the method addToGlobalChi2DerMatrix.
 
      gsl_matrix_set_zero(M1);
      gsl_matrix_set_zero(M2);
      // First, all terms d^2 chi^2/dx1 dx2
      for (auto fo : fitobjects) {
        assert(fo);
        fo->addToGlobalChi2DerMatrix(M1->block->data, M1->tda);
        fo->addToGlobCov(M2->block->data, M2->tda);
      }
      // multiply by -1
      gsl_matrix_scale(M1, -1);
 
      // JL: dy/eta are the derivatives of the "objective function" with respect to the MEASURED parameters.
      // Since the soft constraints do not depend on the measured, but only on the fitted (!) parameters,
      // dy/deta stays -1*M1 also in the presence of soft constraints!
      gsl_matrix_view dydeta  = gsl_matrix_submatrix(M1, 0, 0, idim, npar);
      gsl_matrix_view Cov_eta = gsl_matrix_submatrix(M2, 0, 0, npar, npar);
 
      if (debug > 13) {
        B2INFO("NewFitterGSL::calcCovMatrix\n");
        debug_print(&dydeta.matrix, "dydeta");
        debug_print(&Cov_eta.matrix, "Cov_eta");
      }
 
      // JL: calculates d^2 chi^2 / dx1 dx2 + first (!) derivatives of hard & soft constraints, and the
      // second derivatives of the soft constraints times the values of the fitted parameters
      // - all of the with respect to the FITTED parameters, therefore with soft constraints like in the fit itself.
      assembleM(MatW, vecx, true);
 
      if (debug > 13) {
        debug_print(MatW, "MatW");
      }
 
      // Now, solve M*dadeta = dydeta
 
      // Calculate LU decomposition of M into M3
      int signum;
      int result = gsl_linalg_LU_decomp(MatW, permw, &signum);
 
      if (debug > 13) {
        B2INFO("calcCovMatrix: gsl_linalg_LU_decomp result=" << result);
        debug_print(MatW, "M_LU");
      }
 
      // Calculate inverse of M, store in M3
      gsl_set_error_handler_off();
      int ifail = gsl_linalg_LU_invert(MatW, permw, M3);
      if (ifail) {
        B2WARNING("NewFitterGSL: MatW matrix is singular!");
        return ifail;
      }
 
      if (debug > 13) {
        B2INFO("calcCovMatrix: gsl_linalg_LU_invert ifail=" << ifail);
        debug_print(M3, "Minv");
      }
 
      // Calculate dadeta = M3*dydeta
      gsl_matrix_set_zero(M4);
      gsl_matrix_view dadeta   = gsl_matrix_submatrix(M4, 0, 0, idim, npar);
 
      if (debug > 13) {
        debug_print(&dadeta.matrix, "dadeta");
      }
 
      // dadeta = 1*M*dydeta + 0*dadeta
      gsl_blas_dgemm(CblasNoTrans, CblasNoTrans, 1, M3, &dydeta.matrix, 0, &dadeta.matrix);
 
 
      // Now calculate Cov_a = dadeta*Cov_eta*dadeta^T
 
      // First, calculate M3 = Cov_eta*dadeta^T as
      gsl_matrix_view M3part   = gsl_matrix_submatrix(M3, 0, 0, npar, idim);
      gsl_blas_dgemm(CblasNoTrans, CblasTrans, 1, &Cov_eta.matrix, &dadeta.matrix, 0, &M3part.matrix);
      // Now Cov_a = dadeta*M3part
      gsl_matrix_set_zero(M5);
      gsl_matrix_view  Cov_a = gsl_matrix_submatrix(M5, 0, 0, npar, npar);
      gsl_blas_dgemm(CblasNoTrans, CblasNoTrans, 1, &dadeta.matrix, &M3part.matrix, 0, M5);
      gsl_matrix_memcpy(CCinv, M5);
 
      if (debug > 13) {
        debug_print(&Cov_a.matrix, "Cov_a");
        debug_print(CCinv, "full Cov from err prop");
        debug_print(M1, "uncorr Cov from err prop");
      }
 
      // Finally, copy covariance matrix
      if (cov && covDim != npar) {
        delete[] cov;
        cov = nullptr;
      }
      covDim = npar;
      if (!cov) cov = new double[covDim * covDim];
      for (int i = 0; i < covDim; ++i) {
        for (int j = 0; j < covDim; ++j) {
          cov[i * covDim + j] = gsl_matrix_get(&Cov_a.matrix, i, j);
        }
      }
      covValid = true;
      return 0;
    }

calcLimitedDx()

int calcLimitedDx	(	double &	alpha,
		double &	mu,
		gsl_vector *	vecxnew,
		int	imode,
		gsl_vector *	vecx,
		gsl_vector *	vecdxhat,
		const gsl_vector *	vecdx,
		const gsl_vector *	vecdxscal,
		const gsl_vector *	vece,
		const gsl_matrix *	MatM,
		const gsl_matrix *	MatMscal,
		gsl_matrix *	MatW,
		gsl_vector *	vecw
	)

Parameters

alpha	Output value alpha
mu	Value of mu for merit function
vecxnew	New vector x
imode	mode: 0=Armijo, 1=Wolfe, 2=Goldstein
vecx	Current vector x
vecdxhat	the 2nd order correction step, if any
vecdx	Update vector dx
vecdxscal	Result: Update vector dx, scaled
vece	Error vector e
MatM	Matrix M
MatMscal	Matrix M, scaled
MatW	Work matrix
vecw	Work vector w1

Definition at line 913 of file NewFitterGSL.cc.

    {
      assert(vecxnew);
      assert(vecxnew->size == idim);
      assert(vecx);
      assert(vecx->size == idim);
      assert(vecdxhat);
      assert(vecdxhat->size == idim);
      assert(vecdx);
      assert(vecdx->size == idim);
      assert(vecdxscal);
      assert(vecdxscal->size == idim);
      assert(vece);
      assert(vece->size == idim);
      assert(MatM);
      assert(MatM->size1 == idim && MatM->size2 == idim);
      assert(MatMscal);
      assert(MatMscal->size1 == idim && MatMscal->size2 == idim);
      assert(MatW);
      assert(MatW->size1 == idim && MatW->size2 == idim);
      assert(vecw);
      assert(vecw->size == idim);
 
      alpha = 1;
      double eta = 0.1;
 
      add(vecxnew, vecx, alpha, vecdx);
 
      if (debug > 15) {
        B2INFO("calcLimitedDx: After solving equations: \n");
        debug_print(vecx, "x");
        gsl_blas_dcopy(vecx, vecw);
        gsl_vector_div(vecw, vece);
        debug_print(vecw, "xscal");
        debug_print(vecxnew, "xnew");
        gsl_blas_dcopy(vecxnew, vecw);
        gsl_vector_div(vecw, vece);
        debug_print(vecw, "xnewscal");
      }
 
      mu = calcMu(vecx, vece, vecdx, vecdxscal, vecxnew, MatM, MatMscal, vecw);
 
      updateParams(vecx);
 
      double phi0  = meritFunction(mu, vecx, vece);
      double dphi0 = meritFunctionDeriv(mu, vecx, vece, vecdx, vecw);
 
      if (debug > 12) {
        double eps = 1E-6;
        add(vecw, vecx, eps, vecdx);
        updateParams(vecw);
        double dphinum = (meritFunction(mu, vecw, vece) - phi0) / eps;
        updateParams(vecx);
 
        B2INFO("analytic der: " << dphi0 << ", num=" << dphinum);
      }
 
#ifndef FIT_TRACEOFF
      traceValues["alpha"] = 0;
      traceValues["phi"] = phi0;
      traceValues["mu"] = mu;
      if (tracer) tracer->substep(*this, 0);
#endif
 
      updateParams(vecxnew);
 
      double phiR = meritFunction(mu, vecxnew, vece);
 
      B2DEBUG(12, "calcLimitedDx: phi0=" << phi0 << ", dphi0=" << dphi0
              << ", phiR = " << phiR << ", mu=" << mu
              << ", threshold = " << phi0 + eta * dphi0
              << " => test=" << (phiR > phi0 + eta * dphi0));
 
#ifndef FIT_TRACEOFF
      traceValues["alpha"] = 1;
      traceValues["phi"] = phiR;
      if (tracer) tracer->substep(*this, 0);
#endif
 
 
      // Try Armijo's rule for alpha=1 first, do linesearch only if it fails
      if (phiR > phi0 + eta * alpha * dphi0) {
 
        // try second order correction first
        if (try2ndOrderCorr) {
          calc2ndOrderCorr(vecdxhat, vecxnew, MatM, MatW, vecw);
          gsl_blas_dcopy(vecxnew, vecw);
          add(vecxnew, vecxnew, 1, vecdxhat);
          updateParams(vecxnew);
          double phi2ndOrder  = meritFunction(mu, vecxnew, vece);
 
#ifndef FIT_TRACEOFF
          traceValues["alpha"] = 1.5;
          traceValues["phi"] = phi2ndOrder;
          if (tracer) tracer->substep(*this, 2);
#endif
 
          B2DEBUG(12, "calcLimitedDx: tried 2nd order correction, phi2ndOrder = "
                  << phi2ndOrder << ", threshold = " << phi0 + eta * dphi0);
          if (debug > 15) {
            debug_print(vecdxhat, "dxhat");
            debug_print(xnew, "xnew");
          }
          if (phi2ndOrder <= phi0 + eta * alpha * dphi0) {
            B2DEBUG(12, "  -> 2nd order correction successful!");
            return 1;
          }
          B2DEBUG(12, "  -> 2nd order correction failed, do linesearch!");
          gsl_blas_dcopy(vecw, vecxnew);
          updateParams(vecxnew);
#ifndef FIT_TRACEOFF
          calcChi2();
          traceValues["alpha"] = 1;
          traceValues["phi"] = phiR;
          if (tracer) tracer->substep(*this, 2);
#endif
 
        }
 
        double zeta = 0.5;
        doLineSearch(alpha, vecxnew, imode, phi0, dphi0, eta, zeta, mu,
                     vecx, vecdx, vece, vecw);
      }
 
      return 0;
    }

calcMu()

double calcMu	(	const gsl_vector *	vecx,
		const gsl_vector *	vece,
		const gsl_vector *	vecdx,
		const gsl_vector *	vecdxscal,
		const gsl_vector *	xnew,
		const gsl_matrix *	MatM,
		const gsl_matrix *	MatMscal,
		gsl_vector *	vecw
	)

Parameters

vecx	Current vector x
vece	Current errors x
vecdx	Current step dx
vecdxscal	Current step dx, scaled
xnew	New vector x
MatM	Current matrix M
MatMscal	Current scaled matrix M
vecw	Work vector w

Definition at line 1161 of file NewFitterGSL.cc.

    {
      assert(vecx);
      assert(vecx->size == idim);
      assert(vece);
      assert(vece->size == idim);
      assert(vecdx);
      assert(vecdx->size == idim);
      assert(vecdxscal);
      assert(vecdxscal->size == idim);
      assert(MatM);
      assert(MatM->size1 == idim && MatM->size2 == idim);
      assert(MatMscal);
      assert(MatMscal->size1 == idim && MatMscal->size2 == idim);
      assert(vecw);
      assert(vecw->size == idim);
 
      double result = 0;
      switch (imerit) {
        case 1: { // l1 penalty function, Nocedal&Wright Eq. (15.24)
          result = 0;
//         for (int kglobal = npar; kglobal < npar+ncon; ++kglobal) {
//           double abslambda = std::fabs (gsl_vector_get (vecx, kglobal));
//           if (abslambda > result)
//             result = abslambda;
//         }
          // calculate grad f^T*p
          gsl_vector_set_zero(vecw);
          addConstraints(vecw);
          gsl_vector_view c(gsl_vector_subvector(vecw, npar, ncon));
          // ||c||_1
          double cnorm1 = gsl_blas_dasum(&c.vector);
          // scale constraint values by 1/(delta e)
          gsl_vector_const_view lambdaerr(gsl_vector_const_subvector(vece, npar, ncon));
          gsl_vector_mul(&c.vector, &lambdaerr.vector);
          // ||c||_1
          double cnorm1scal = gsl_blas_dasum(&c.vector);
 
          double rho = 0.1;
          double eps = 0.001;
 
          assembleChi2Der(vecw);
          gsl_vector_view gradf(gsl_vector_subvector(vecw, 0, npar));
          gsl_vector_const_view p(gsl_vector_const_subvector(vecdx, 0, npar));
          double gradfTp;
          gsl_blas_ddot(&gradf.vector, &p.vector, &gradfTp);
 
          B2DEBUG(17, "NewFitterGSL::calcMu: cnorm1scal=" << cnorm1scal
                  << ", gradfTp=" << gradfTp);
 
          // all constraints very well fulfilled, use max(lambda+1) criterium
          if (cnorm1scal < ncon * eps || gradfTp <= 0) {
            for (int kglobal = npar; kglobal < npar + ncon; ++kglobal) {
              double abslambda = std::fabs(gsl_vector_get(vecxnew, kglobal));
              if (abslambda > result)
                result = abslambda;
            }
            result /= (1 - rho);
          } else {
            // calculate p^T L p
            double pTLp = calcpTLp(vecdx, MatM, vecw);
            double sigma = (pTLp > 0) ? 1 : 0;
            B2DEBUG(17, "  pTLp = " << pTLp);
            // Nocedal&Wright Eq. (18.36)
            result = (gradfTp + 0.5 * sigma * pTLp) / ((1 - rho) * cnorm1);
          }
          B2DEBUG(17, "  result = " << result);
        }
        break;
        case 2: // l1 penalty function, errors scaled, Nocedal&Wright Eq. (15.24)
          result = 0;
          for (int kglobal = npar; kglobal < npar + ncon; ++kglobal) {
            double abslambdascal = std::fabs(gsl_vector_get(vecx, kglobal) / gsl_vector_get(vece, kglobal));
            if (abslambdascal > result)
              result = abslambdascal;
          }
          break;
        default: assert(0);
      }
      return result;
 
    }

calcNewtonDx()

int calcNewtonDx	(	gsl_vector *	vecdx,
		gsl_vector *	vecdxscal,
		gsl_vector *	vecx,
		const gsl_vector *	vece,
		gsl_matrix *	MatM,
		gsl_matrix *	MatMscal,
		gsl_vector *	vecy,
		gsl_vector *	vecyscal,
		gsl_matrix *	MatW,
		gsl_matrix *	MatW2,
		gsl_permutation *	permW,
		gsl_vector *	vecw
	)

Parameters

vecdx	Result: Update vector dx
vecdxscal	Result: Update vector dx, scaled
vecx	Current vector x
vece	Current `‘error’' set of x
MatM	Matrix M
MatMscal	Matrix M, scaled
vecy	Vector y
vecyscal	Vector y, scaled,
MatW	Work matrix
MatW2	Work matrix
permW	Work permutation vector
vecw	Work vector

Definition at line 802 of file NewFitterGSL.cc.

    {
      assert(vecdx);
      assert(vecdx->size == idim);
      assert(vecdxscal);
      assert(vecdxscal->size == idim);
      assert(vecx);
      assert(vecx->size == idim);
      assert(vece);
      assert(vece->size == idim);
      assert(MatM);
      assert(MatM->size1 == idim && MatM->size2 == idim);
      assert(MatMscal);
      assert(MatMscal->size1 == idim && MatMscal->size2 == idim);
      assert(vecy);
      assert(vecy->size == idim);
      assert(vecyscal);
      assert(vecyscal->size == idim);
      assert(MatW);
      assert(MatW->size1 == idim && MatW->size2 == idim);
      assert(MatW2);
      assert(MatW2->size1 == idim && MatW2->size2 == idim);
      assert(permw);
      assert(permw->size == idim);
      assert(vecw);
      assert(vecw->size == idim);
 
      int ncalc = 0;
      double ptLp = 0;
 
      do {
        if (ncalc == 1) {
          // try to recalculate lambdas
          assembleConstDer(MatM);
          determineLambdas(vecx, MatM, vecx, MatW, vecw);
          B2DEBUG(12, "NewFitterGSL::calcNewtonDx: ptLp=" << ptLp << " with lambdas from last iteration");
        } else if (ncalc == 2) {
          // try to set lambdas to zero
          gsl_vector_view lambda(gsl_vector_subvector(vecx, npar, ncon));
          gsl_vector_set_zero(&lambda.vector);
          B2DEBUG(12, "NewFitterGSL::calcNewtonDx: ptLp=" << ptLp << " with recalculated lambdas");
        } else if (ncalc >= 3) {
          B2DEBUG(12, "NewFitterGSL::calcNewtonDx: ptLp=" << ptLp << " with zero lambdas");
          break;
        }
 
        if (debug > 15) {
          B2INFO("calcNewtonDx: before setting up equations: \n");
          debug_print(vecx, "x");
        }
 
        assembleM(MatM, vecx);
        if (!isfinite(MatM)) return 1;
 
        scaleM(MatMscal, MatM, vece);
        assembley(vecy, vecx);
        if (!isfinite(vecy)) return 2;
        scaley(vecyscal, vecy, vece);
 
        if (debug > 15) {
          B2INFO("calcNewtonDx: After setting up equations: \n");
          debug_print(MatM, "M");
          debug_print(MatMscal, "Mscal");
          debug_print(vecy, "y");
          debug_print(vecyscal, "yscal");
          debug_print(vece, "perr");
          debug_print(vecx, "x");
          debug_print(xnew, "xnew");
        }
 
        // from x_(n+1) = x_n - y/y' = x_n - M^(-1)*y we have M*(x_n-x_(n+1)) = y,
        // which we solve for dx = x_n-x_(n+1) and hence x_(n+1) = x_n-dx
 
        double epsLU = 1E-12;
        double epsSV = 1E-3;
        double detW;
        solveSystem(vecdxscal, detW, vecyscal, MatMscal, MatW, MatW2, vecw, epsLU, epsSV);
 
 
#ifndef FIT_TRACEOFF
        traceValues["detW"] = detW;
#endif
 
        // step is - computed vector
        gsl_blas_dscal(-1, dxscal);
 
        // dx = dxscal*e (component wise)
        gsl_vector_memcpy(vecdx, vecdxscal);
        gsl_vector_mul(vecdx, vece);
 
        if (debug > 15) {
          B2INFO("calcNewtonDx: Result: \n");
          debug_print(vecdx, "dx");
          debug_print(vecdxscal, "dxscal");
        }
 
 
        ptLp = calcpTLp(dx, M, v1);
        ++ncalc;
      } while (ptLp < 0);
 
      return 0;
    }

calcpTLp()

double calcpTLp	(	const gsl_vector *	vecdx,
		const gsl_matrix *	MatM,
		gsl_vector *	vecw
	)

Parameters

vecdx	Current step dx
MatM	Current matrix M
vecw	Work vector w

Definition at line 1613 of file NewFitterGSL.cc.

    {
      assert(vecdx);
      assert(vecdx->size == idim);
      assert(MatM);
      assert(MatM->size1 == idim && MatM->size2 == idim);
      assert(vecw);
      assert(vecw->size == idim);
 
      gsl_vector_const_view p(gsl_vector_const_subvector(vecdx, 0, npar));
      gsl_vector_view Lp(gsl_vector_subvector(vecw, 0, npar));
      gsl_matrix_const_view L(gsl_matrix_const_submatrix(MatM, 0, 0, npar, npar));
      gsl_blas_dsymv(CblasUpper, 1, &L.matrix, &p.vector, 0, &Lp.vector);
      double result;
      gsl_blas_ddot(&p.vector, &Lp.vector, &result);
 
      return result;
    }

calcReducedHessian()

gsl_matrix_view calcReducedHessian ( int &  rankH, gsl_matrix *  MatW1, const gsl_vector *  vecx, gsl_matrix *  MatW2, gsl_matrix *  MatW3, gsl_vector *  vecw1, gsl_vector *  vecw2, gsl_permutation *  permW, double  eps = 0  )

virtual

Calculate reduced Hessian; return value is a matrix view of MatW1 giving the reduced Hessian.

Parameters

rankH	dimension of H
MatW1	work matrix, contains H at the end
vecx	vector with current x values
MatW2	work matrix, contains Z at the end
MatW3	work matrix
vecw1	work vector
vecw2	work vector
permW	Work permutation vector
eps	Singular values < eps*(max(abs(s_i))) are set to 0

Definition at line 1870 of file NewFitterGSL.cc.

    {
      assert(MatW1);
      assert(MatW1->size1 == idim && MatW1->size2 == idim);
      assert(vecx);
      assert(vecx->size == idim);
      assert(MatW2);
      assert(MatW2->size1 == idim && MatW2->size2 == idim);
      assert(MatW3);
      assert(MatW2->size1 == idim && MatW2->size2 == idim);
      assert(vecw1);
      assert(vecw1->size == idim);
      assert(vecw2);
      assert(vecw2->size == idim);
      assert(permw);
      assert(permw->size == idim);
 
      int rankA;
      // Z is a matrix view of MatW2!
      gsl_matrix_view Z = calcZ(rankA, MatW2, MatW1, vecw1, vecw2, permw, eps);
 
      // fill Lagrangian
      assembleG(MatW1, vecx);
 
      rankH =  npar - rankA;
 
      gsl_matrix_view G(gsl_matrix_submatrix(MatW1, 0,    0,    npar,  npar));
      gsl_matrix_view GZ(gsl_matrix_submatrix(MatW3, 0,    0,    npar,  rankH));
      gsl_matrix_view ZGZ(gsl_matrix_submatrix(MatW1, 0,    0,    rankH, rankH));
 
      // Calculate Z^T G Z
 
      // GZ = 1*G*Z + 0*GZ
      gsl_blas_dsymm(CblasLeft, CblasUpper, 1, &G.matrix, &Z.matrix, 0, &GZ.matrix);
      // ZGZ = 1*Z^T*GZ + 0*ZGZ
      gsl_blas_dgemm(CblasTrans, CblasNoTrans, 1, &Z.matrix, &GZ.matrix, 0, &ZGZ.matrix);
 
      return ZGZ;
    }

calcReducedHessianEigenvalues()

gsl_vector_view calcReducedHessianEigenvalues ( int &  rankH, gsl_matrix *  MatW1, const gsl_vector *  vecx, gsl_matrix *  MatW2, gsl_matrix *  MatW3, gsl_vector *  vecw1, gsl_vector *  vecw2, gsl_permutation *  permW, gsl_eigen_symm_workspace *  eigenws, double  eps = 0  )

virtual

Parameters

rankH	dimension of H
MatW1	work matrix, contains H at the end
vecx	vector with current x values
MatW2	work matrix, contains Z at the end
MatW3	work matrix
vecw1	work vector
vecw2	work vector
permW	Work permutation vector
eigenws	Work space for eigenvalue calculation
eps	Singular values < eps*(max(abs(s_i))) are set to 0

Definition at line 1914 of file NewFitterGSL.cc.

    {
      assert(MatW1);
      assert(MatW1->size1 == idim && MatW1->size2 == idim);
      assert(vecx);
      assert(vecx->size == idim);
      assert(MatW2);
      assert(MatW2->size1 == idim && MatW2->size2 == idim);
      assert(MatW3);
      assert(MatW2->size1 == idim && MatW2->size2 == idim);
      assert(vecw1);
      assert(vecw1->size == idim);
      assert(vecw2);
      assert(vecw2->size == idim);
      assert(permw);
      assert(permw->size == idim);
      assert(eigensws);
 
      gsl_matrix_view Hred = calcReducedHessian(rankH, MatW1, vecx, MatW2, MatW3, vecw1, vecw2, permw, eps);
 
      gsl_matrix_view Hredcopy(gsl_matrix_submatrix(MatW3, 0,    0,    rankH, rankH));
      // copy Hred -> Hredcopy
      gsl_matrix_memcpy(&Hredcopy.matrix, &Hred.matrix);
 
      gsl_vector_view eval(gsl_vector_subvector(vecw1, 0, rankH));
      gsl_eigen_symm(&Hredcopy.matrix, &eval.vector, eigensws);
 
      return eval;
    }

calcZ()

gsl_matrix_view calcZ ( int &  rankA, gsl_matrix *  MatW1, gsl_matrix *  MatW2, gsl_vector *  vecw1, gsl_vector *  vecw2, gsl_permutation *  permW, double  eps = 0  )

virtual

Calculate null space of constraints; return value is a matrix view of MatW1, with column vectors spanning null(A)

Parameters

rankA	rank of A (= number of lin. indep. constraints)
MatW1	work matrix, contains Z at the end
MatW2	work matrix
vecw1	work vector
vecw2	work vector
permW	Work permutation vector
eps	Singular values < eps*(max(abs(s_i))) are set to 0

Definition at line 1834 of file NewFitterGSL.cc.

    {
      assert(MatW1);
      assert(MatW1->size1 == idim && MatW1->size2 == idim);
      assert(MatW2);
      assert(MatW2->size1 == idim && MatW2->size2 == idim);
      assert(vecw1);
      assert(vecw1->size == idim);
      assert(vecw2);
      assert(vecw2->size == idim);
      assert(permw);
      assert(permw->size == idim);
 
      // fill A and AT
      assembleConstDer(MatW2);
 
      gsl_matrix_const_view AT(gsl_matrix_const_submatrix(MatW2, 0,    0,    npar, npar));
      //gsl_matrix_view QR       (gsl_matrix_submatrix       (MatW2, 0,    0,    npar, npar));
      gsl_matrix_view Q(gsl_matrix_submatrix(MatW1, 0,    0,    npar, npar));
      gsl_matrix_view R(gsl_matrix_submatrix(MatW1, npar, 0,    ncon, npar));
 
      int signum = 0;
      //gsl_linalg_QRPT_decomp   (&QR.matrix, vecw1, permw, &signum, vecw2);
      //gsl_linalg_QR_unpack     (&QR.matrix, vecw1, &Q.matrix, &R.matrix);
      gsl_linalg_QRPT_decomp2(&AT.matrix, &Q.matrix, &R.matrix, vecw1, permw, &signum, vecw2);
 
      rankA = 0;
      for (int i = 0; i < ncon; ++i) {
        if (fabs(gsl_matrix_get(&R.matrix, i, i)) > eps) rankA++;
      }
      auto result = gsl_matrix_view(gsl_matrix_submatrix(MatW1, 0, rankA, ncon - rankA, npar));
      return result;
    }

debug_print() [1/2]

void debug_print ( const gsl_matrix *  m, const char *  name  )

static

Definition at line 470 of file NewFitterGSL.cc.

    {
      for (unsigned int  i = 0; i < m->size1; ++i)
        for (unsigned int j = 0; j < m->size2; ++j)
          if (gsl_matrix_get(m, i, j) != 0)
            B2INFO(name << "[" << i << "][" << j << "]=" << gsl_matrix_get(m, i, j));
    }

debug_print() [2/2]

void debug_print ( const gsl_vector *  v, const char *  name  )

static

Definition at line 478 of file NewFitterGSL.cc.

    {
      for (unsigned int  i = 0; i < v->size; ++i)
        if (gsl_vector_get(v, i) != 0)
          B2INFO(name << "[" << i << "]=" << gsl_vector_get(v, i));
    }

determineLambdas()

int determineLambdas ( gsl_vector *  vecxnew, const gsl_matrix *  MatM, const gsl_vector *  vecx, gsl_matrix *  MatW, gsl_vector *  vecw, double  eps = 0  )

virtual

Determine best lambda values.

Parameters

vecxnew	vector with new lambda values
MatM	matrix with constraint derivatives
vecx	vector with current x values
MatW	work matrix
vecw	work vector
eps	Singular values < eps*(max(abs(s_i))) are set to 0

Definition at line 1482 of file NewFitterGSL.cc.

    {
      assert(vecxnew);
      assert(vecxnew->size == idim);
      assert(MatM);
      assert(MatM->size1 == idim && MatM->size2 == idim);
      assert(vecx);
      assert(vecx->size == idim);
      assert(MatW);
      assert(MatW->size1 == idim && MatW->size2 == idim);
      assert(vecw);
      assert(vecw->size == idim);
      assert(idim == static_cast<unsigned int>(npar + ncon));
 
      gsl_matrix_const_view A(gsl_matrix_const_submatrix(MatM, 0, npar, npar, ncon));
      gsl_matrix_view ATA(gsl_matrix_submatrix(MatW, npar, npar, ncon, ncon));
      gsl_matrix_view Acopy(gsl_matrix_submatrix(MatW, 0, npar, npar, ncon));
      gsl_matrix_view& V = ATA;
 
      gsl_vector_view gradf(gsl_vector_subvector(vecw, 0, npar));
      gsl_vector_view ATgradf(gsl_vector_subvector(vecw, npar, ncon));
      gsl_vector_view& s = ATgradf;
 
      gsl_vector_view lambdanew(gsl_vector_subvector(vecxnew, npar, ncon));
 
      if (debug > 15) {
//        gsl_vector_const_view lambda(gsl_vector_const_subvector(vecx, npar, ncon));
        B2INFO("lambda: ");
        gsl_vector_fprintf(stdout, &lambdanew.vector, "%f");
      }
 
      // ATA = 1*A^T*A + 0*ATA
      gsl_blas_dgemm(CblasTrans, CblasNoTrans, 1, &A.matrix, &A.matrix, 0, &ATA.matrix);
 
      // put grad(f) into vecw
      int isfail = assembleChi2Der(vecw);
      if (isfail) return isfail;
 
 
      // ATgradf = -1*A^T*gradf + 0*ATgradf
      gsl_blas_dgemv(CblasTrans, -1, &A.matrix, &gradf.vector, 0, &ATgradf.vector);
 
      if (debug > 17) {
        B2INFO("A: ");
        gsl_matrix_fprintf(stdout, &A.matrix, "%f");
        B2INFO("ATA: ");
        gsl_matrix_fprintf(stdout, &ATA.matrix, "%f");
        B2INFO("gradf: ");
        gsl_vector_fprintf(stdout, &gradf.vector, "%f");
        B2INFO("ATgradf: ");
        gsl_vector_fprintf(stdout, &ATgradf.vector, "%f");
      }
 
      // solve ATA * lambdanew = ATgradf using the Cholsky factorization method
      gsl_error_handler_t* old_handler =  gsl_set_error_handler_off();
      int cholesky_result = gsl_linalg_cholesky_decomp(&ATA.matrix);
      gsl_set_error_handler(old_handler);
      if (cholesky_result) {
        B2INFO("NewFitterGSL::determineLambdas: resorting to SVD");
        // ATA is not positive definite, i.e. A does not have full column rank
        // => use the SVD of A to solve A lambdanew = gradf
 
        // Acopy = A
        gsl_matrix_memcpy(&Acopy.matrix, &A.matrix);
 
        // SVD decomposition of Acopy
        gsl_linalg_SV_decomp_jacobi(&Acopy.matrix, &V.matrix, &s.vector);
        // set small values to zero
        double mins = eps * std::fabs(gsl_vector_get(&s.vector, 0));
        for (int i = 0; i < ncon; ++i) {
          if (std::fabs(gsl_vector_get(&s.vector, i)) <= mins)
            gsl_vector_set(&s.vector, i, 0);
        }
        gsl_linalg_SV_solve(&Acopy.matrix, &V.matrix, &s.vector, &gradf.vector, &lambdanew.vector);
      } else {
        gsl_linalg_cholesky_solve(&ATA.matrix, &ATgradf.vector, &lambdanew.vector);
      }
      if (debug > 15) {
        B2INFO("lambdanew: ");
        gsl_vector_fprintf(stdout, &lambdanew.vector, "%f");
      }
      return 0;
    }

doLineSearch()

int doLineSearch	(	double &	alpha,
		gsl_vector *	vecxnew,
		int	imode,
		double	phi0,
		double	dphi0,
		double	eta,
		double	zeta,
		double	mu,
		const gsl_vector *	vecx,
		const gsl_vector *	vecdx,
		const gsl_vector *	vece,
		gsl_vector *	vecw
	)

Parameters

alpha	Output value alpha
vecxnew	New vector x
imode	mode: 0=Armijo, 1=Wolfe, 2=Goldstein
phi0	Merit function for alpha=0
dphi0	Directional derivative of merit function for alpha=0
eta	Constant for Armijo's rule
zeta	Constant for Wolfe's or Goldstein's rule
mu	Value of mu for merit function
vecx	Current vector x
vecdx	Update vector dx
vece	Error vector e
vecw	Work vector w

Definition at line 1047 of file NewFitterGSL.cc.

    {
      assert(vecxnew);
      assert(vecxnew->size == idim);
      assert(vecx);
      assert(vecx->size == idim);
      assert(vecdx);
      assert(vecdx->size == idim);
      assert(vece);
      assert(vece->size == idim);
      assert(vecw);
      assert(vecw->size == idim);
 
      // don't do anything for imode = -1
      if (imode < 0) return -1;
 
      assert((imode == 0) || eta < zeta);
 
      if (dphi0 >= 0) {
        // Difficult situation: merit function will increase,
        // thus every step makes it worse
        // => choose the minimum step and return
        B2DEBUG(12, "NewFitterGSL::doLineSearch: dphi0 > 0!");
        // Choose new alpha
        alpha = 0.001;
 
        add(vecxnew, vecx, alpha, vecdx);
        updateParams(vecxnew);
 
        double phi = meritFunction(mu, vecxnew, vece);
 
#ifndef FIT_TRACEOFF
        traceValues["alpha"] = alpha;
        traceValues["phi"] = phi;
        if (tracer) tracer->substep(*this, 1);
#endif
        return 2;
      }
 
      // alpha=1 already tried
      double alphaR = alpha;
      // vecxnew = vecx + alpha*vecdx
      add(vecxnew, vecx, alpha, vecdx);
      updateParams(vecxnew);
 
      double alphaL = 0;
      //  double phiL = phi0;   //fix warning
      //  double dphiL = dphi0; //fix warning
 
      double dphi;
      int nite = 0;
 
      do {
        nite++;
        // Choose new alpha
        alpha = 0.5 * (alphaL + alphaR);
 
        add(vecxnew, vecx, alpha, vecdx);
        updateParams(vecxnew);
 
        double phi = meritFunction(mu, vecxnew, vece);
 
#ifndef FIT_TRACEOFF
        traceValues["alpha"] = alpha;
        traceValues["phi"] = phi;
        if (tracer) tracer->substep(*this, 1);
#endif
 
        // Armijo's rule always holds
        if (phi >= phi0 + eta * alpha * dphi0) {
          B2DEBUG(15, "NewFitterGSL::doLineSearch, Armijo: phi=" << phi
                  << " >= " << phi0 + eta * alpha * dphi0
                  << " at alpha=" << alpha);
          alphaR = alpha;
          //      phiR = phi;   //fix warning
          continue;
        }
 
        // imode == 0 (Armijo) asserted above
        if (imode == 1) { // Wolfe
          dphi = meritFunctionDeriv(mu, vecxnew, vece, vecdx, vecw);
          if (dphi < zeta * dphi0) {
            B2DEBUG(15, "NewFitterGSL::doLineSearch, Wolfe: dphi=" << dphi
                    << " < " << zeta * dphi0
                    << " at alpha=" << alpha);
            alphaL = alpha;
            //        phiL   = phi; //fix warning
            //        dphiL  = dphi; //fix warning
          } else {
            break;
          }
        } else {                // Goldstein
          if (phi < phi0 + zeta * alpha * dphi0) {
            B2DEBUG(15, "NewFitterGSL::doLineSearch, Goldstein: phi=" << phi
                    << " < " << phi0 + zeta * alpha * dphi0
                    << " at alpha=" << alpha);
            alphaL = alpha;
            //        phiL   = phi; //fix warning
          } else {
            break;
          }
        }
      } while (nite < 30 && (alphaL == 0 || nite < 6));
      if (alphaL > 0) alpha = alphaL;
      return 1;
    }

fillperr()

void fillperr

(

gsl_vector *

vece

)

Definition at line 556 of file NewFitterGSL.cc.

    {
      assert(vece);
      assert(vece->size == idim);
      gsl_vector_set_all(vece, 1);
      for (auto fo : fitobjects) {
        assert(fo);
        for (int ilocal = 0; ilocal < fo->getNPar(); ++ilocal) {
          if (!fo->isParamFixed(ilocal)) {
            int iglobal = fo->getGlobalParNum(ilocal);
            assert(iglobal >= 0 && iglobal < npar);
            double e = std::abs(fo->getError(ilocal));
            gsl_vector_set(vece, iglobal, e ? e : 1);
          }
        }
      }
      for (auto c : constraints) {
        assert(c);
        int iglobal = c->getGlobalNum();
        assert(iglobal >= 0 && iglobal < (int)idim);
        double e =  c->getError();
        gsl_vector_set(vece, iglobal, e ? 1 / e : 1);
      }
    }

fillx()

void fillx

(

gsl_vector *

vecx

)

Definition at line 538 of file NewFitterGSL.cc.

    {
      assert(vecx);
      assert(vecx->size == idim);
 
      gsl_vector_set_zero(vecx);
      for (auto fo : fitobjects) {
        assert(fo);
        for (int ilocal = 0; ilocal < fo->getNPar(); ++ilocal) {
          if (!fo->isParamFixed(ilocal)) {
            int iglobal = fo->getGlobalParNum(ilocal);
            assert(iglobal >= 0 && iglobal < npar);
            gsl_vector_set(vecx, iglobal, fo->getParam(ilocal));
          }
        }
      }
    }

fit()

double fit ( )

overridevirtual

The fit method, returns the fit probability.

Implements BaseFitter.

Definition at line 143 of file NewFitterGSL.cc.

    {
 
      // order parameters etc
      initialize();
 
      // initialize eta, etasv, y
      assert(x && (x->size == idim));
      assert(xold && (xold->size == idim));
      assert(xnew && (xnew->size == idim));
      assert(dx && (dx->size == idim));
      assert(y && (y->size == idim));
      assert(perr && (perr->size == idim));
      assert(v1 && (v1->size == idim));
      assert(v2 && (v2->size == idim));
//   assert (Meval && Meval->size == idim);
      assert(M && (M->size1 == idim && M->size2 == idim));
      assert(W && (W->size1 == idim && W->size2 == idim));
      assert(W2 && (W2->size1 == idim && W2->size2 == idim));
      assert(M1 && (M1->size1 == idim && M1->size2 == idim));
//   assert (Mevec && Mevec->size1 == idim && Mevec->size2 == idim);
      assert(permW && (permW->size == idim));
 
      gsl_vector_set_zero(x);
      gsl_vector_set_zero(y);
      gsl_vector_set_all(perr, 1);
 
      // Store initial x values in x
      fillx(x);
      if (debug > 14) {
        B2INFO("fit: Start values: \n");
        debug_print(x, "x");
      }
      // make sure parameters are consistent
      updateParams(x);
      fillx(x);
 
      assembleConstDer(M);
      int ifailL = determineLambdas(x, M, x, W, v1);
      if (ifailL) return -1;
 
      // Get starting values into x
//  gsl_vector_memcpy (x, xold);
 
 
      // LET THE GAMES BEGIN
 
#ifndef FIT_TRACEOFF
      calcChi2();
      traceValues["alpha"] = 0;
      traceValues["phi"] = 0;
      traceValues["mu"] = 0;
      traceValues["detW"] = 0;
      if (tracer) tracer->initialize(*this);
#endif
 
      bool converged = false;
      ierr = 0;
 
      chi2new = calcChi2();
      nit = 0;
 
      do {
#ifndef FIT_TRACEOFF
        if (tracer) tracer->step(*this);
#endif
 
        chi2old = chi2new;
        // Store old x values in xold
        gsl_blas_dcopy(x, xold);
        // Fill errors into perr
        fillperr(perr);
 
        // Now, calculate the result vector y with the values of the derivatives
        // d chi^2/d x
        int ifail = calcNewtonDx(dx, dxscal, x, perr, M, Mscal, y, yscal, W, W2, permW, v1);
 
        if (ifail) {
          ierr = 99;
          B2DEBUG(10, "NewFitterGSL::fit: calcNewtonDx error " << ifail);
 
          break;
        }
 
        if (debug > 15) {
          B2INFO("Before test convergence, calcNewtonDe: Result: \n");
          debug_print(dx, "dx");
          debug_print(dxscal, "dxscal");
        }
 
        // test convergence:
        if (gsl_blas_dasum(dxscal) < 1E-6 * idim) {
          converged = true;
          break;
        }
 
        double alpha = 1;
        double mu = 0;
        int imode = 2;
 
        calcLimitedDx(alpha, mu, xnew, imode, x, v2, dx, dxscal, perr, M, Mscal, W, v1);
 
        gsl_blas_dcopy(xnew, x);
 
        chi2new = calcChi2();
 
//   *-- Convergence criteria
 
        ++nit;
        if (nit > 200) ierr = 1;
 
        converged = (abs(chi2new - chi2old) < 0.0001);
 
//     if (abs (chi2new - chi2old) >= 0.001)
//       B2INFO("abs (chi2new - chi2old)=" << abs (chi2new - chi2old) << " -> try again\n");
//     if (fvalbest >= 1E-3)
//       B2INFO("fvalbest=" << fvalbest << " -> try again\n");
//     if (fvalbest >= 1E-6 && abs(fvals[0]-fvalbest) >= 0.2*fvalbest )
//       B2INFO("fvalbest=" << fvalbest
//            << ", abs(fvals[0]-fvalbest)=" << abs(fvals[0]-fvalbest)<< " -> try again\n");
//     if (stepbest >= 1E-3)
//       B2INFO("stepbest=" << stepbest << " -> try again\n");
//     B2INFO("converged=" << converged );
        if (converged) {
          B2DEBUG(12, "abs (chi2new - chi2old)=" << abs(chi2new - chi2old) << "\n"
                  << "fvalbest=" << fvalbest << "\n"
                  << "abs(fvals[0]-fvalbest)=" << abs(fvals[0] - fvalbest) << "\n");
        }
 
      } while (!(converged || ierr));
 
 
#ifndef FIT_TRACEOFF
      if (tracer) tracer->step(*this);
#endif
 
//*-- End of iterations - calculate errors.
 
// ERROR CALCULATION
 
      if (!ierr) {
 
        int ifailw = calcCovMatrix(W, permW, x);
 
        if (!ifailw) {
          // update errors in fitobjects
          for (auto& fitobject : fitobjects) {
            for (int ilocal = 0; ilocal < fitobject->getNPar(); ++ilocal) {
              int iglobal = fitobject->getGlobalParNum(ilocal);
              for (int jlocal = ilocal; jlocal < fitobject->getNPar(); ++jlocal) {
                int jglobal = fitobject->getGlobalParNum(jlocal);
                if (iglobal >= 0 && jglobal >= 0)
                  fitobject->setCov(ilocal, jlocal, gsl_matrix_get(CCinv, iglobal, jglobal));
              }
            }
          }
        } else ierr = 999;
      }
 
 
      if (debug > 11) {
        B2INFO("========= END =========\n");
        B2INFO("Fit objects:\n");
        for (auto fo : fitobjects) {
          assert(fo);
          B2INFO(fo->getName() << ": " << *fo << ", chi2=" << fo->getChi2());
        }
        B2INFO("constraints:\n");
        for (auto i = constraints.begin(); i != constraints.end(); ++i) {
          BaseHardConstraint* c = *i;
          assert(c);
          B2INFO(i - constraints.begin() << " " << c->getName() << ": " << c->getValue() << "+-" << c->getError());
        }
        B2INFO("=============================================\n");
      }
 
// *-- Turn chisq into probability.
      fitprob = (chi2new >= 0 && ncon + nsoft - nunm > 0) ? gsl_cdf_chisq_Q(chi2new, ncon + nsoft - nunm) : -1;
 
#ifndef FIT_TRACEOFF
      if (tracer) tracer->finish(*this);
#endif
 
      B2DEBUG(10, "NewFitterGSL::fit: converged=" << converged
              << ", nit=" << nit << ", fitprob=" << fitprob);
 
      if (ierr > 0) fitprob = -1;
 
      return fitprob;
 
    }

getChi2()

double getChi2 ( ) const

overridevirtual

Get the chi**2 of the last fit.

Implements BaseFitter.

Definition at line 432 of file NewFitterGSL.cc.

{return chi2;}

getConstraints()

std::vector< BaseHardConstraint * > * getConstraints ( )

virtualinherited

Definition at line 112 of file BaseFitter.cc.

    {
      return &constraints;
    }

getDoF()

int getDoF ( ) const

overridevirtual

Get the number of degrees of freedom of the last fit.

Implements BaseFitter.

Definition at line 433 of file NewFitterGSL.cc.

{return ncon + nsoft - nunm;}

getError()

int getError ( ) const

overridevirtual

Get the error code of the last fit: 0=OK, 1=failed.

Implements BaseFitter.

Definition at line 430 of file NewFitterGSL.cc.

{return ierr;}

getFitObjects()

std::vector< BaseFitObject * > * getFitObjects ( )

virtualinherited

Definition at line 107 of file BaseFitter.cc.

    {
      return &fitobjects;
    }

getGlobalCovarianceMatrix() [1/2]

double * getGlobalCovarianceMatrix ( int &  idim )

virtualinherited

Parameters

idim	1st dimension of global covariance matrix

Definition at line 158 of file BaseFitter.cc.

    {
      if (covValid && cov) {
        idim = covDim;
        return cov;
      }
      idim = 0;
      return nullptr;
    }

getGlobalCovarianceMatrix() [2/2]

const double * getGlobalCovarianceMatrix ( int &  idim ) const

virtualinherited

Parameters

idim	1st dimension of global covariance matrix

Definition at line 148 of file BaseFitter.cc.

    {
      if (covValid && cov) {
        idim = covDim;
        return cov;
      }
      idim = 0;
      return nullptr;
    }

getIterations()

int getIterations ( ) const

overridevirtual

Get the number of iterations of the last fit.

Implements BaseFitter.

Definition at line 434 of file NewFitterGSL.cc.

{return nit;}

getNcon()

int getNcon ( ) const

virtual

Get the number of hard constraints of the last fit.

Definition at line 497 of file NewFitterGSL.cc.

{return ncon;}

getNpar()

int getNpar ( ) const

virtual

Get the number of all parameters of the last fit.

Definition at line 500 of file NewFitterGSL.cc.

{return npar;}

getNsoft()

int getNsoft ( ) const

virtual

Get the number of soft constraints of the last fit.

Definition at line 498 of file NewFitterGSL.cc.

{return nsoft;}

getNunm()

int getNunm ( ) const

virtual

Get the number of unmeasured parameters of the last fit.

Definition at line 499 of file NewFitterGSL.cc.

{return nunm;}

getProbability()

double getProbability ( ) const

overridevirtual

Get the fit probability of the last fit.

Implements BaseFitter.

Definition at line 431 of file NewFitterGSL.cc.

{return fitprob;}

getSoftConstraints()

std::vector< BaseSoftConstraint * > * getSoftConstraints ( )

virtualinherited

Definition at line 117 of file BaseFitter.cc.

    {
      return &softconstraints;
    }

getTracer() [1/2]

BaseTracer * getTracer ( )

virtualinherited

Definition at line 130 of file BaseFitter.cc.

    {
      return tracer;
    }

getTracer() [2/2]

const BaseTracer * getTracer ( ) const

virtualinherited

Definition at line 134 of file BaseFitter.cc.

    {
      return tracer;
    }

ini_gsl_matrix()

void ini_gsl_matrix ( gsl_matrix *&  m, int unsigned  size1, unsigned int  size2  )

static

Definition at line 459 of file NewFitterGSL.cc.

    {
      if (m) {
        if (m->size1 != size1 || m->size2 != size2) {
          gsl_matrix_free(m);
          if (size1 * size2 > 0) m = gsl_matrix_alloc(size1, size2);
          else m = nullptr;
        }
      } else if (size1 * size2 > 0) m = gsl_matrix_alloc(size1, size2);
    }

ini_gsl_permutation()

void ini_gsl_permutation ( gsl_permutation *&  p, unsigned int  size  )

static

Definition at line 436 of file NewFitterGSL.cc.

    {
      if (p) {
        if (p->size != size) {
          gsl_permutation_free(p);
          if (size > 0) p = gsl_permutation_alloc(size);
          else p = nullptr;
        }
      } else if (size > 0) p = gsl_permutation_alloc(size);
    }

ini_gsl_vector()

void ini_gsl_vector ( gsl_vector *&  v, int unsigned  size  )

static

Definition at line 447 of file NewFitterGSL.cc.

    {
 
      if (v) {
        if (v->size != size) {
          gsl_vector_free(v);
          if (size > 0) v = gsl_vector_alloc(size);
          else v = nullptr;
        }
      } else if (size > 0) v = gsl_vector_alloc(size);
    }

initialize()

bool initialize ( )

overridevirtual

Initialize the fitter.

Implements BaseFitter.

Definition at line 335 of file NewFitterGSL.cc.

    {
      covValid = false;
//  bool debug = true;
 
      // tell fitobjects the global ordering of their parameters:
      npar = 0;
      nunm = 0;
      //
      for (auto& fitobject : fitobjects) {
        for (int ilocal = 0; ilocal < fitobject->getNPar(); ++ilocal) {
          if (!fitobject->isParamFixed(ilocal)) {
            fitobject->setGlobalParNum(ilocal, npar);
            ++npar;
            if (!fitobject->isParamMeasured(ilocal)) ++nunm;
          }
        }
      }
 
      // set number of constraints
      ncon = constraints.size();
      // Tell the constraints their numbers
      for (unsigned int icon = 0; icon < constraints.size(); ++icon) {
        BaseHardConstraint* c = constraints[icon];
        assert(c);
        c->setGlobalNum(npar + icon);
      }
 
      // JL: should soft constraints have numbers assigned as well?
      nsoft = softconstraints.size();
 
      if (nunm > ncon + nsoft) {
        B2ERROR("NewFitterGSL::initialize: nunm=" << nunm << " > ncon+nsoft="
                << ncon << "+" << nsoft);
      }
 
      // dimension of "big M" matrix
      idim = npar + ncon;
 
      ini_gsl_vector(x, idim);
      ini_gsl_vector(xold, idim);
      ini_gsl_vector(xnew, idim);
//   ini_gsl_vector (xbest, idim);
      ini_gsl_vector(dx, idim);
      ini_gsl_vector(dxscal, idim);
//   ini_gsl_vector (grad, idim);
      ini_gsl_vector(y, idim);
      ini_gsl_vector(yscal, idim);
      ini_gsl_vector(perr, idim);
      ini_gsl_vector(v1, idim);
      ini_gsl_vector(v2, idim);
//   ini_gsl_vector (Meval, idim);
 
      ini_gsl_matrix(M, idim, idim);
      ini_gsl_matrix(Mscal, idim, idim);
      ini_gsl_matrix(W, idim, idim);
      ini_gsl_matrix(W2, idim, idim);
      ini_gsl_matrix(W3, idim, idim);
      ini_gsl_matrix(M1, idim, idim);
      ini_gsl_matrix(M2, idim, idim);
      ini_gsl_matrix(M3, idim, idim);
      ini_gsl_matrix(M4, idim, idim);
      ini_gsl_matrix(M5, idim, idim);
//   ini_gsl_matrix (Mevec, idim, idim);
      ini_gsl_matrix(CC, idim, idim);
      ini_gsl_matrix(CC1, idim, idim);
      ini_gsl_matrix(CCinv, idim, idim);
 
      ini_gsl_permutation(permW, idim);
 
      if (eigenws && eigenwsdim != idim) {
        gsl_eigen_symm_free(eigenws);
        eigenws = nullptr;
      }
      if (eigenws == nullptr) eigenws = gsl_eigen_symm_alloc(idim);
      eigenwsdim = idim;
 
      return true;
 
    }

invertM()

int invertM

(

)

Definition at line 1342 of file NewFitterGSL.cc.

    {
      gsl_matrix_memcpy(W, M);
 
      int ifail = 0;
 
      int signum;
      // Calculate LU decomposition of M into W
      int result = gsl_linalg_LU_decomp(W, permW, &signum);
      B2DEBUG(11, "invertM: gsl_linalg_LU_decomp result=" << result);
      // Calculate inverse of M
      ifail = gsl_linalg_LU_invert(W, permW, M);
      B2DEBUG(11, "invertM: gsl_linalg_LU_invert result=" << ifail);
 
      if (ifail != 0) {
        B2ERROR("NewtonFitter::invert: ifail from gsl_linalg_LU_invert=" << ifail);
      }
      return ifail;
    }

isfinite() [1/2]

bool isfinite ( const gsl_matrix *  mat )

static

Definition at line 528 of file NewFitterGSL.cc.

    {
      if (mat == nullptr) return true;
      for (size_t i = 0; i < mat->size1; ++i)
        for (size_t j = 0; j < mat->size2; ++j)
          if (!std::isfinite(gsl_matrix_get(mat, i, j))) return false;
      return true;
    }

isfinite() [2/2]

bool isfinite ( const gsl_vector *  vec )

static

Definition at line 520 of file NewFitterGSL.cc.

    {
      if (vec == nullptr) return true;
      for (size_t i = 0; i < vec->size; ++i)
        if (!std::isfinite(gsl_vector_get(vec, i))) return false;
      return true;
    }

meritFunction()

double meritFunction	(	double	mu,
		const gsl_vector *	vecx,
		const gsl_vector *	vece
	)

Parameters

mu	Value of mu
vecx	Current vector x
vece	Current errors x

Definition at line 1249 of file NewFitterGSL.cc.

    {
      assert(vecx);
      assert(vecx->size == idim);
      assert(vece);
      assert(vece->size == idim);
 
      double result = 0;
      switch (imerit) {
        case 1: // l1 penalty function, Nocedal&Wright Eq. (15.24)
          result = calcChi2();
          for (auto c : constraints) {
            assert(c);
            int kglobal = c->getGlobalNum();
            assert(kglobal >= 0 && kglobal < (int)idim);
            result += mu * std::fabs(c->getValue());
          }
          break;
        case 2: // l1 penalty function, errors scaled, Nocedal&Wright Eq. (15.24)
          result = calcChi2();
          for (auto c : constraints) {
            assert(c);
            int kglobal = c->getGlobalNum();
            assert(kglobal >= 0 && kglobal < (int)idim);
            // vece[kglobal] is 1/error for constraint k
            result += mu * std::fabs(c->getValue() * gsl_vector_get(vece, kglobal));
          }
          break;
        default: assert(0);
      }
      return result;
 
    }

meritFunctionDeriv()

double meritFunctionDeriv	(	double	mu,
		const gsl_vector *	vecx,
		const gsl_vector *	vece,
		const gsl_vector *	vecdx,
		gsl_vector *	vecw
	)

Parameters

mu	Value of mu
vecx	Current vector x
vece	Current errors x
vecdx	Current update vector dx
vecw	work vector

Definition at line 1284 of file NewFitterGSL.cc.

    {
      assert(vecx);
      assert(vecx->size == idim);
      assert(vece);
      assert(vece->size == idim);
      assert(vecdx);
      assert(vecdx->size == idim);
      assert(vecw);
      assert(vecw->size == idim);
 
      double result = 0;
      switch (imerit) {
        case 1: // l1 penalty function, Nocedal&Wright Eq. (15.24), Eq. (18.29)
          assembleChi2Der(vecw);
          for (int i = 0; i < npar; ++i) {
            result += gsl_vector_get(vecdx, i) * gsl_vector_get(vecw, i);
          }
//       for (ConstraintIterator i = constraints.begin(); i != constraints.end(); ++i) {
//         BaseHardConstraint *c = *i;
//         assert (c);
//         int kglobal = c->getGlobalNum();
//         assert (kglobal >= 0 && kglobal < (int)idim);
//         result +=  c->dirDerAbs (vecdx->block->data, vecw->block->data, npar, mu);
//       }
          for (auto c : constraints) {
            assert(c);
            int kglobal = c->getGlobalNum();
            assert(kglobal >= 0 && kglobal < (int)idim);
            result -=  mu * std::fabs(c->getValue());
          }
          break;
        case 2: // l1 penalty function, errors scaled, Nocedal&Wright Eq. (15.24), Eq. (18.29)
          assembleChi2Der(vecw);
          for (int i = 0; i < npar; ++i) {
            result += gsl_vector_get(vecdx, i) * gsl_vector_get(vecw, i);
          }
//       for (ConstraintIterator i = constraints.begin(); i != constraints.end(); ++i) {
//         BaseHardConstraint *c = *i;
//         assert (c);
//         int kglobal = c->getGlobalNum();
//         assert (kglobal >= 0 && kglobal < (int)idim);
//         result +=  c->dirDerAbs (vecdx->block->data, vecw->block->data, npar, mu*gsl_vector_get (vece, kglobal));
//       }
          for (auto c : constraints) {
            assert(c);
            int kglobal = c->getGlobalNum();
            assert(kglobal >= 0 && kglobal < (int)idim);
            result -=  mu * std::fabs(c->getValue()) * gsl_vector_get(vece, kglobal);
          }
          break;
        default: assert(0);
      }
      return result;
    }

MoorePenroseInverse()

void MoorePenroseInverse ( gsl_matrix *  Ainv, gsl_matrix *  A, gsl_matrix *  W, gsl_vector *  w, double  eps = 0  )

static

Compute the Moore-Penrose pseudo-inverse A+ of A, using SVD.

Parameters

Ainv	Result: m x n matrix A+
A	Input: n x m matrix A, n >= m (is destroyed!)
W	Work matrix, at least m x m
w	Work vector w, at least m
eps	Singular values < eps*(max(abs(s_i))) are set to 0

Definition at line 1569 of file NewFitterGSL.cc.

    {
      assert(Ainv);
      assert(A);
      assert(Ainv->size1 == A->size2 && Ainv->size2 == A->size1);
      assert(A->size1 >= A->size2);
      assert(Wm);
      assert(Wm->size1 >= A->size1 && Wm->size2 >= A->size2);
      assert(w);
      assert(w->size >= A->size2);
 
      int n = A->size1;
      int m = A->size2;
 
      // Original A -> A diag(w) Wm^T
      gsl_linalg_SV_decomp_jacobi(A, Wm, w);
 
      double mins = eps * std::fabs(gsl_vector_get(w, 0));
 
      // Compute pseudo-inverse of diag(w)
      for (int i = 0; i < m; ++i) {
        double singval = gsl_vector_get(w, i);
        if (std::fabs(singval) > mins)
          gsl_vector_set(w, i, 1 / singval);
        else
          gsl_vector_set(w, i, 0);
      }
 
      // Compute Ainv = Wm diag(w) A^T
 
      // first: Ainv = Wm* diag(w)
      for (int j = 0; j < n; ++j) {
        double wval = gsl_vector_get(w, j);
        for (int i = 0; i < m; ++i)
          gsl_matrix_set(Wm, i, j, wval * gsl_matrix_get(Wm, i, j));
      }
      // Ainv = 1*Wm*A^T + 0*Ainv
      gsl_blas_dgemm(CblasNoTrans, CblasTrans, 1, Wm, A, 0, Ainv);
 
    }

reset()

void reset ( )

virtualinherited

Definition at line 122 of file BaseFitter.cc.

    {
      fitobjects.resize(0);
      constraints.resize(0);
      softconstraints.resize(0);
      covValid = false;
    }

scaleM()

void scaleM	(	gsl_matrix *	MatMscal,
		const gsl_matrix *	MatM,
		const gsl_vector *	vece
	)

Definition at line 687 of file NewFitterGSL.cc.

    {
      assert(MatMscal);
      assert(MatMscal->size1 == idim && MatMscal->size2 == idim);
      assert(MatM);
      assert(MatM->size1 == idim && MatM->size2 == idim);
      assert(vece);
      assert(vece->size == idim);
 
      // Rescale columns and rows by perr
      for (unsigned int i = 0; i < idim; ++i)
        for (unsigned int j = 0; j < idim; ++j)
          gsl_matrix_set(MatMscal, i, j,
                         gsl_vector_get(vece, i)*gsl_vector_get(vece, j)*gsl_matrix_get(MatM, i, j));
    }

scaley()

void scaley	(	gsl_vector *	vecyscal,
		const gsl_vector *	vecy,
		const gsl_vector *	vece
	)

Definition at line 737 of file NewFitterGSL.cc.

    {
      assert(vecyscal);
      assert(vecyscal->size == idim);
      assert(vecy);
      assert(vecy->size == idim);
      assert(vece);
      assert(vece->size == idim);
 
      gsl_vector_memcpy(vecyscal, vecy);
      gsl_vector_mul(vecyscal, vece);
    }

setDebug()

void setDebug ( int  debuglevel )

virtual

Set the Debug Level.

Definition at line 1362 of file NewFitterGSL.cc.

    {
      debug = Debuglevel;
    }

setTracer() [1/2]

void setTracer ( BaseTracer &  newTracer )

virtualinherited

Definition at line 143 of file BaseFitter.cc.

    {
      tracer = &newTracer;
    }

setTracer() [2/2]

void setTracer ( BaseTracer *  newTracer )

virtualinherited

Definition at line 138 of file BaseFitter.cc.

    {
      tracer = newTracer;
    }

solveSystem()

int solveSystem	(	gsl_vector *	vecdxscal,
		double &	detW,
		const gsl_vector *	vecyscal,
		const gsl_matrix *	MatMscal,
		gsl_matrix *	MatW,
		gsl_matrix *	MatW2,
		gsl_vector *	vecw,
		double	epsLU,
		double	epsSV
	)

solve system of equations Mscal*dxscal = yscal

Definition at line 1704 of file NewFitterGSL.cc.

    {
      assert(vecdxscal);
      assert(vecdxscal->size == idim);
      assert(vecyscal);
      assert(vecyscal->size == idim);
      assert(MatMscal);
      assert(MatMscal->size1 == idim && MatMscal->size2 == idim);
      assert(MatW);
      assert(MatW->size1 == idim && MatW->size2 == idim);
      assert(MatW2);
      assert(MatW2->size1 == idim && MatW2->size2 == idim);
      assert(vecw);
      assert(vecw->size == idim);
 
      int result = 0;
 
      int iLU = solveSystemLU(vecdxscal, detW, vecyscal, MatMscal, MatW, vecw, epsLU);
      if (iLU == 0) return result;
 
      result = 1;
      int iSVD = solveSystemSVD(vecdxscal, vecyscal, MatMscal, MatW, MatW2, vecw, epsSV);
      if (iSVD == 0) return result;
 
      return -1;
    }

solveSystemLU()

int solveSystemLU	(	gsl_vector *	vecdxscal,
		double &	detW,
		const gsl_vector *	vecyscal,
		const gsl_matrix *	MatMscal,
		gsl_matrix *	MatW,
		gsl_vector *	vecw,
		double	eps
	)

solve system of equations Mscal*dxscal = yscal using LU decomposition

Definition at line 1739 of file NewFitterGSL.cc.

    {
      assert(vecdxscal);
      assert(vecdxscal->size == idim);
      assert(vecyscal);
      assert(vecyscal->size == idim);
      assert(MatMscal);
      assert(MatMscal->size1 == idim && MatMscal->size2 == idim);
      assert(MatW);
      assert(MatW->size1 == idim && MatW->size2 == idim);
      assert(vecw);
      assert(vecw->size == idim);
 
      gsl_matrix_memcpy(MatW, MatMscal);
 
      int ifail = 0;
      detW = 0;
 
      int signum;
      int result = gsl_linalg_LU_decomp(MatW, permW, &signum);
      B2DEBUG(14, "NewFitterGSL::solveSystem: gsl_linalg_LU_decomp result=" << result);
      if (result != 0) return 1;
 
      detW = gsl_linalg_LU_det(MatW, signum);
      B2DEBUG(14, "NewFitterGSL::solveSystem: determinant of W=" << detW);
      if (std::fabs(detW) < eps) return 2;
      if (!std::isfinite(detW)) {
        B2DEBUG(10, "NewFitterGSL::solveSystem: infinite determinant of W=" << detW);
        return 3;
      }
      if (debug > 15) {
        B2INFO("NewFitterGSL::solveSystem: after LU decomposition: \n");
        debug_print(MatW, "W");
      }
      // Solve W*dxscal = yscal
      ifail = gsl_linalg_LU_solve(MatW, permW, vecyscal, vecdxscal);
      B2DEBUG(14, "NewFitterGSL::solveSystem: gsl_linalg_LU_solve result=" << ifail);
 
      if (ifail != 0) {
        B2ERROR("NewFitterGSL::solveSystem: ifail from gsl_linalg_LU_solve=" << ifail);
        return 3;
      }
      return 0;
    }

solveSystemSVD()

int solveSystemSVD	(	gsl_vector *	vecdxscal,
		const gsl_vector *	vecyscal,
		const gsl_matrix *	MatMscal,
		gsl_matrix *	MatW,
		gsl_matrix *	MatW2,
		gsl_vector *	vecw,
		double	eps
	)

solve system of equations Mscal*dxscal = yscal using SVD decomposition

Definition at line 1793 of file NewFitterGSL.cc.

    {
      assert(vecdxscal);
      assert(vecdxscal->size == idim);
      assert(vecyscal);
      assert(vecyscal->size == idim);
      assert(MatMscal);
      assert(MatMscal->size1 == idim && MatMscal->size2 == idim);
      assert(MatW);
      assert(MatW->size1 == idim && MatW->size2 == idim);
      assert(MatW2);
      assert(MatW2->size1 == idim && MatW2->size2 == idim);
      assert(vecw);
      assert(vecw->size == idim);
 
      B2DEBUG(10,  "solveSystemSVD called");
 
      gsl_matrix_memcpy(MatW, MatMscal);
 
      // SVD decomposition of MatW
      gsl_linalg_SV_decomp_jacobi(MatW, MatW2, vecw);
      // set small values to zero
      double mins = eps * std::fabs(gsl_vector_get(vecw, 0));
      B2DEBUG(15,  "SV 0 = " << gsl_vector_get(vecw, 0));
      for (unsigned int i = 0; i < idim; ++i) {
        if (std::fabs(gsl_vector_get(vecw, i)) <= mins) {
          B2DEBUG(15, "Setting SV" << i << " = " << gsl_vector_get(vecw, i) << " to zero!");
          gsl_vector_set(vecw, i, 0);
        }
      }
      gsl_linalg_SV_solve(MatW, MatW2, vecw, vecyscal, vecdxscal);
      return 0;
    }

updateParams()

bool updateParams

(

gsl_vector *

vecx

)

Definition at line 502 of file NewFitterGSL.cc.

    {
      assert(vecx);
      assert(vecx->size == idim);
      bool significant = false;
      for (auto i = fitobjects.begin(); i != fitobjects.end(); ++i) {
        BaseFitObject* fo = *i;
        assert(fo);
        bool s = fo->updateParams(vecx->block->data, vecx->size);
        significant |=  s;
        if (nit < nitdebug && s) {
          B2DEBUG(15, "Significant update for FO " << i - fitobjects.begin() << " ("
                  << fo->getName() << ")\n");
        }
      }
      return significant;
    }

Member Data Documentation

CC

gsl_matrix* CC

Definition at line 353 of file NewFitterGSL.h.

CC1

gsl_matrix* CC1

Definition at line 354 of file NewFitterGSL.h.

CCinv

gsl_matrix* CCinv

Definition at line 355 of file NewFitterGSL.h.

chi2

double chi2

final chi2

Definition at line 262 of file NewFitterGSL.h.

chi2best

double chi2best

Definition at line 362 of file NewFitterGSL.h.

chi2new

double chi2new

Definition at line 363 of file NewFitterGSL.h.

chi2old

double chi2old

Definition at line 364 of file NewFitterGSL.h.

constraints

ConstraintContainer constraints

protectedinherited

Definition at line 100 of file BaseFitter.h.

cov

double* cov

protectedinherited

global covariance matrix of last fit problem

Definition at line 104 of file BaseFitter.h.

covDim

int covDim

protectedinherited

dimension of global covariance matrix

Definition at line 103 of file BaseFitter.h.

covValid

bool covValid

protectedinherited

Flag whether global covariance is valid.

Definition at line 105 of file BaseFitter.h.

debug

int debug

Definition at line 377 of file NewFitterGSL.h.

dx

gsl_vector* dx

Definition at line 331 of file NewFitterGSL.h.

dxscal

gsl_vector* dxscal

Definition at line 332 of file NewFitterGSL.h.

eigenws

gsl_eigen_symm_workspace* eigenws

Definition at line 358 of file NewFitterGSL.h.

eigenwsdim

unsigned int eigenwsdim

Definition at line 360 of file NewFitterGSL.h.

fitobjects

FitObjectContainer fitobjects

protectedinherited

Definition at line 99 of file BaseFitter.h.

fitprob

double fitprob

fit probability

Definition at line 261 of file NewFitterGSL.h.

fvalbest

double fvalbest

Definition at line 365 of file NewFitterGSL.h.

fvals

double fvals[NITMAX]

Definition at line 372 of file NewFitterGSL.h.

idim

unsigned int idim

Definition at line 326 of file NewFitterGSL.h.

ierr

int ierr

Error status.

Definition at line 258 of file NewFitterGSL.h.

imerit

int imerit

Definition at line 374 of file NewFitterGSL.h.

M

gsl_matrix* M

Definition at line 341 of file NewFitterGSL.h.

M1

gsl_matrix* M1

Definition at line 347 of file NewFitterGSL.h.

M2

gsl_matrix* M2

Definition at line 348 of file NewFitterGSL.h.

M3

gsl_matrix* M3

Definition at line 349 of file NewFitterGSL.h.

M4

gsl_matrix* M4

Definition at line 350 of file NewFitterGSL.h.

M5

gsl_matrix* M5

Definition at line 351 of file NewFitterGSL.h.

Mscal

gsl_matrix* Mscal

Definition at line 342 of file NewFitterGSL.h.

ncon

int ncon

total number of hard constraints

Definition at line 255 of file NewFitterGSL.h.

nit

int nit

Number of iterations.

Definition at line 259 of file NewFitterGSL.h.

npar

int npar

total number of parameters

Definition at line 254 of file NewFitterGSL.h.

nsoft

int nsoft

total number of soft constraints

Definition at line 256 of file NewFitterGSL.h.

nunm

int nunm

total number of unmeasured parameters

Definition at line 257 of file NewFitterGSL.h.

permW

gsl_permutation* permW

Definition at line 357 of file NewFitterGSL.h.

perr

gsl_vector* perr

Definition at line 336 of file NewFitterGSL.h.

scale

double scale

Definition at line 366 of file NewFitterGSL.h.

scalebest

double scalebest

Definition at line 367 of file NewFitterGSL.h.

scalevals

double scalevals[NITMAX]

Definition at line 371 of file NewFitterGSL.h.

softconstraints

SoftConstraintContainer softconstraints

protectedinherited

Definition at line 101 of file BaseFitter.h.

stepbest

double stepbest

Definition at line 369 of file NewFitterGSL.h.

stepsize

double stepsize

Definition at line 368 of file NewFitterGSL.h.

tracer

BaseTracer* tracer

protectedinherited

Definition at line 108 of file BaseFitter.h.

traceValues

std::map<std::string, double> traceValues

inherited

Definition at line 112 of file BaseFitter.h.

try2ndOrderCorr

bool try2ndOrderCorr

Definition at line 375 of file NewFitterGSL.h.

v1

gsl_vector* v1

Definition at line 337 of file NewFitterGSL.h.

v2

gsl_vector* v2

Definition at line 338 of file NewFitterGSL.h.

W

gsl_matrix* W

Definition at line 343 of file NewFitterGSL.h.

W2

gsl_matrix* W2

Definition at line 344 of file NewFitterGSL.h.

W3

gsl_matrix* W3

Definition at line 345 of file NewFitterGSL.h.

x

gsl_vector* x

Definition at line 327 of file NewFitterGSL.h.

xnew

gsl_vector* xnew

Definition at line 329 of file NewFitterGSL.h.

xold

gsl_vector* xold

Definition at line 328 of file NewFitterGSL.h.

y

gsl_vector* y

Definition at line 334 of file NewFitterGSL.h.

yscal

gsl_vector* yscal

Definition at line 335 of file NewFitterGSL.h.

---

The documentation for this class was generated from the following files:

• analysis/OrcaKinFit/include/NewFitterGSL.h
• analysis/OrcaKinFit/src/NewFitterGSL.cc