A class for correlated Gaussian observables. More...

#include <CorrelatedGaussianObservables.h>

Detailed Description

A class for correlated Gaussian observables.

Author: HEPfit Collaboration

Copyright: GNU General Public License

This class builds the correlated Gaussian observables that are specified in the SomeModel.conf file or specified by the user.

Definition at line 34 of file CorrelatedGaussianObservables.h.

Public Member Functions
void	AddObs (Observable &Obs_i)
	A method to add observables to the list of correlated Gaussian observables. More...

void	ComputeCov (const TMatrixDSym &Corr)
	Computes the covariance matrix for the correlated Gaussian observables set. More...

std::vector< double >	computeLeaveOneOutWeights ()
	A method to compute the full weight and the weights associated with the observable when one observable is left out of the set. More...

virtual double	computeWeight ()
	A method to compute the weight associated with the observable. More...

	CorrelatedGaussianObservables ()
	The default Constructor. More...

	CorrelatedGaussianObservables (const CorrelatedGaussianObservables &orig)
	The copy constructor. More...

	CorrelatedGaussianObservables (std::string name_i)
	Constructor. More...

const TMatrixDSym &	getCov () const
	A get method to access the covariance matrix of the correlated Gaussian observables. More...

std::string	getName () const
	A get method to access the name of the correlated Gaussian observables set. More...

std::vector< Observable >	getObs () const
	A get method to access the vector of observables that are defined in one correlated Gaussian observables set. More...

Observable	getObs (int i) const
	A get method to access an element of the vector of observables that are defined in one correlated Gaussian observables set. More...

bool	isEOF ()
	A method to check if the end of file has been reached. More...

bool	isPrediction ()
	A method to check if the Correlated Observables are set for prediction. More...

int	ParseCGO (boost::ptr_vector< Observable > &Observables, std::ifstream &ifile, boost::tokenizer< boost::char_separator< char > >::iterator &beg, std::string &infilename, ThObsFactory &myObsFactory, StandardModel *myModel, int lineNo, int rank)
	The parser for CorrelatedGaussianObservables. More...

void	setCovarianceFromConfig (bool setInvCov)
	A method to specify whether the inverse covariance is being set from the config file. More...

void	setIsPrediction (bool IsPrediction_i)
	A method to set a set of CGO to be predicted. More...

void	setName (std::string name_i)
	A method to set a set of CGO to be predicted. More...

virtual	~CorrelatedGaussianObservables ()
	The default destructor. More...

Private Attributes
bool	covarianceFromConfig

std::string	filepath
	The path to the config file being parsed. More...

TMatrixDSym	InvCov
	The inverse covariance matrix. More...

bool	IsEOF
	A boolean which is true if the end of file is reached. More...

bool	IsPrediction
	Flag to define a set of Correlated Observables to be predicted. More...

std::string	name
	The name of the correlated Gaussian Observables set. More...

std::vector< Observable >	Obs
	A vector of observables whose correlation will be calculated. More...

TVectorD	x
	The vector of the differences between the theory and average values of the observables in the set. More...

Constructor & Destructor Documentation

◆ CorrelatedGaussianObservables() [1/3]

CorrelatedGaussianObservables::CorrelatedGaussianObservables ( std::string name_i )

Constructor.

Parameters

[in] name_i a given name for the set of correlated Gaussian observables

Definition at line 15 of file CorrelatedGaussianObservables.cpp.

                                                                             {
    name = name_i;
    IsPrediction = false;
    IsEOF = false;
    covarianceFromConfig = false;
}

◆ CorrelatedGaussianObservables() [2/3]

CorrelatedGaussianObservables::CorrelatedGaussianObservables ( )

The default Constructor.

Definition at line 22 of file CorrelatedGaussianObservables.cpp.

                                                             {
    name = "";
    IsPrediction = false;
    IsEOF = false;
    covarianceFromConfig = false;
}

◆ CorrelatedGaussianObservables() [3/3]

CorrelatedGaussianObservables::CorrelatedGaussianObservables ( const CorrelatedGaussianObservables & orig )

The copy constructor.

Definition at line 29 of file CorrelatedGaussianObservables.cpp.

                                                                                                      : InvCov(orig.InvCov), x(orig.x) {
    Obs = orig.Obs;
    name = orig.name;
    IsPrediction = orig.IsPrediction;
    IsEOF = orig.IsEOF;
    covarianceFromConfig = false;
}

◆ ~CorrelatedGaussianObservables()

CorrelatedGaussianObservables::~CorrelatedGaussianObservables ( )

virtual

The default destructor.

Definition at line 37 of file CorrelatedGaussianObservables.cpp.

37 {

38}

Member Function Documentation

◆ AddObs()

void CorrelatedGaussianObservables::AddObs ( Observable & Obs_i )

A method to add observables to the list of correlated Gaussian observables.

Parameters

Obs_i reference to an object of type Observable

Definition at line 40 of file CorrelatedGaussianObservables.cpp.

                                                            {
    Obs.push_back(Obs_i);
}

◆ ComputeCov()

void CorrelatedGaussianObservables::ComputeCov ( const TMatrixDSym & Corr )

Computes the covariance matrix for the correlated Gaussian observables set.

Parameters

Corr	the correlation matrix for the correlated Gassian observables set

Definition at line 44 of file CorrelatedGaussianObservables.cpp.

                                                                      {
    unsigned int size = Obs.size();
    if (Corr.GetNrows() != size)
        throw std::runtime_error("The size of the correlated observables in " + name + " does not match the size of the correlation matrix!");
    InvCov.ResizeTo(size, size);
    x.ResizeTo(size);
    if (covarianceFromConfig) {
        for (unsigned int i = 0; i < size; i++) {
            for (unsigned int j = 0; j < size; j++)
                InvCov(i, j) = Corr(i, j);
        }
 
        // Check inverse-covariance is positive semi-definite   
        TMatrixDSymEigen icovES(InvCov);
        TVectorD egval(icovES.GetEigenValues());
        unsigned int EVbad = 0;
        for (unsigned int i = 0; i < size; i++) {
            if (egval(i) < 0.) {
                EVbad++;
            }
        }
        if (EVbad > 0) {
            std::cout << "WARNING: Inverse-covariance matrix of the correlated observables in "<< name <<" is not a positive semi-definite matrix!" << std::endl;
            std::cout << "("<< EVbad <<" non positive eigenvalue(s).)" << std::endl;
            sleep(2);
        }
        
    } else {
        for (unsigned int i = 0; i < size; i++) {
            for (unsigned int j = 0; j < size; j++)
                InvCov(i, j) = Obs.at(i).getErrg() * Corr(i, j) * Obs.at(j).getErrg();            
        }
       
        // Check covariance is positive definite
        TMatrixDSymEigen covES(InvCov);
        TVectorD egval(covES.GetEigenValues());
        unsigned int EVbad = 0;
        for (unsigned int i = 0; i < size; i++) {
            if (egval(i) <= 0.) {
                EVbad++;
            }
        }
        if (EVbad > 0) {
            std::cout << "WARNING: Covariance matrix of the correlated observables in "<< name <<" is not a positive definite matrix!" << std::endl;
            std::cout << "("<< EVbad <<" non positive eigenvalue(s).)" << std::endl;
            sleep(2);
        }
        
        // Invert
        InvCov.Invert();
    }
}

◆ computeLeaveOneOutWeights()

std::vector< double > CorrelatedGaussianObservables::computeLeaveOneOutWeights ( )

A method to compute the full weight and the weights associated with the observable when one observable is left out of the set.

Definition at line 105 of file CorrelatedGaussianObservables.cpp.

                                                                           {
    std::vector<double> weights;
 
    for (unsigned int i = 0; i < Obs.size(); i++) 
        x(i) = Obs.at(i).computeTheoryValue() - Obs.at(i).getAve();
    
    weights.push_back(-0.5 * x * (InvCov * x));
    
    for (unsigned int i = 0; i < Obs.size(); i++) {
        double xi_backup = x(i);
        x(i) = 0.0;
        weights.push_back(-0.5 * x * (InvCov * x));
        x(i) = xi_backup;
    }
 
    return weights;
}

◆ computeWeight()

double CorrelatedGaussianObservables::computeWeight ( )

virtual

A method to compute the weight associated with the observable.

Definition at line 97 of file CorrelatedGaussianObservables.cpp.

                                                    {
        
    for (unsigned int i = 0; i < Obs.size(); i++)
        x(i) = Obs.at(i).computeTheoryValue() - Obs.at(i).getAve();
 
    return (-0.5 * x * (InvCov * x));
}

◆ getCov()

const TMatrixDSym & CorrelatedGaussianObservables::getCov ( ) const

inline

A get method to access the covariance matrix of the correlated Gaussian observables.

Definition at line 112 of file CorrelatedGaussianObservables.h.

    {
        return InvCov;
    }

◆ getName()

std::string CorrelatedGaussianObservables::getName ( ) const

inline

A get method to access the name of the correlated Gaussian observables set.

Returns: the name

Definition at line 104 of file CorrelatedGaussianObservables.h.

    {
        return name;
    }

◆ getObs() [1/2]

std::vector< Observable > CorrelatedGaussianObservables::getObs ( ) const

inline

A get method to access the vector of observables that are defined in one correlated Gaussian observables set.

Returns: a vector of type Observable()

Definition at line 85 of file CorrelatedGaussianObservables.h.

    {
        return Obs;
    }

◆ getObs() [2/2]

Observable CorrelatedGaussianObservables::getObs ( int i ) const

inline

A get method to access an element of the vector of observables that are defined in one correlated Gaussian observables set.

Returns: an element of the vector of type Observable()

Definition at line 95 of file CorrelatedGaussianObservables.h.

    {
        return (Obs.at(i));
    }

◆ isEOF()

bool CorrelatedGaussianObservables::isEOF ( )

inline

A method to check if the end of file has been reached.

Returns: a boolean which is true if the end of file has been reached

Definition at line 140 of file CorrelatedGaussianObservables.h.

    {
        return IsEOF;
    }

◆ isPrediction()

bool CorrelatedGaussianObservables::isPrediction ( )

inline

A method to check if the Correlated Observables are set for prediction.

Returns: a boolean which is true if the set of CGO are to be predicted

Definition at line 158 of file CorrelatedGaussianObservables.h.

    {
        return IsPrediction;
    }

◆ ParseCGO()

int CorrelatedGaussianObservables::ParseCGO	(	boost::ptr_vector< Observable > &	Observables,
		std::ifstream &	ifile,
		boost::tokenizer< boost::char_separator< char > >::iterator &	beg,
		std::string &	infilename,
		ThObsFactory &	myObsFactory,
		StandardModel *	myModel,
		int	lineNo,
		int	rank
	)

The parser for CorrelatedGaussianObservables.

Parameters

[in]	Observables	the pointer vector containing the Observables
[in]	ifile	the stream containing the config file to be parsed
[in]	beg	the iterator that parses a line in the config file
[in]	infilename	the name of the config file being parsed
[in]	myModel	a pointer to the model
[in]	lineNo	the current line number at which the file is being parsed
[in]	rank	the rank of the process that is using the parser

Returns: the line number (integer) after the parsing is done

Definition at line 123 of file CorrelatedGaussianObservables.cpp.

                  {
    if (infilename.find("\\/") == std::string::npos) filepath = infilename.substr(0, infilename.find_last_of("\\/") + 1);
    boost::char_separator<char> sep(" \t");
    name = *beg;
    ++beg;
    int size = atoi((*beg).c_str());
 
    int nlines = 0;
    std::vector<bool> lines;
    std::string line;
    for (int i = 0; i < size; i++) {
        IsEOF = getline(ifile, line).eof();
        if (line.empty() || line.at(0) == '#') {
            if (rank == 0) throw std::runtime_error("ERROR: no comments or empty lines in CorrelatedGaussianObservables please! In file " + infilename + " at line number:" + boost::lexical_cast<std::string>(lineNo) + ".\n");
            else sleep(2);
        }
        lineNo++;
        boost::tokenizer<boost::char_separator<char> > tok(line, sep);
        beg = tok.begin();
        std::string type = *beg;
        ++beg;
        if (type.compare("Observable") != 0 && type.compare("BinnedObservable") != 0 && type.compare("FunctionObservable") != 0) {
            if (rank == 0) throw std::runtime_error("ERROR: in line no." + boost::lexical_cast<std::string>(lineNo) + " of file " + infilename + ", expecting an Observable or BinnedObservable or FunctionObservable type here...\n");
            else sleep(2);
        }
        Observable * tmpObs = new Observable();
        tmpObs->setHasInverseCovariance(covarianceFromConfig);
        beg = tmpObs->ParseObservable(type, &tok, beg, filepath, infilename, rank);
        tmpObs->setTho(myObsFactory.CreateThMethod(tmpObs->getThname(), *myModel));
        if (!IsPrediction) {
            if (tmpObs->isTMCMC()) {
                AddObs(*tmpObs);
                lines.push_back(true);
                nlines++;
            } else {
                Observables.push_back(tmpObs);
                lines.push_back(false);
            }
        } else {
            if (tmpObs->isTMCMC()) {
                if (rank == 0) throw std::runtime_error("ERROR: in line no." + boost::lexical_cast<std::string>(lineNo) + " of file " + infilename + ", Observable/BinnedObservable cannot be set to MCMC in CorrelatedObservable. Use CorrelatedGaussianObservable instead.\n");
                else sleep(2);
            } else {
                AddObs(*tmpObs);
                nlines++;
            }
            if (tmpObs->getDistr().compare("weight") == 0 || tmpObs->getDistr().compare("file") == 0) {
                if (rank == 0) throw std::runtime_error("ERROR: in line no." + boost::lexical_cast<std::string>(lineNo) + " of file " + infilename + ", Observable/BinnedObservable cannot be set to weight or file in CorrelatedObservable. Use CorrelatedGaussianObservable instead.\n");
                else sleep(2);
            }
        }
    }
 
    if (nlines > 1) {
        if (!IsPrediction) {
            TMatrixD myCorr(nlines, nlines);
            int ni = 0;
            for (int i = 0; i < size; i++) {
                IsEOF = getline(ifile, line).eof();
                if (line.empty() || line.at(0) == '#') {
                    if (rank == 0) throw std::runtime_error("ERROR: no comments or empty lines in CorrelatedGaussianObservables please! In file " + infilename + " at line number:" + boost::lexical_cast<std::string>(lineNo) + ".\n");
                    else sleep(2);
                }
                lineNo++;
                if (lines.at(i)) {
                    boost::tokenizer<boost::char_separator<char> > mytok(line, sep);
                    beg = mytok.begin();
                    int nj = 0;
                    for (int j = 0; j < size; j++) {
                        if ((*beg).compare(0, 1, "0") == 0
                                || (*beg).compare(0, 1, "1") == 0
                                || (*beg).compare(0, 1, "-") == 0 || (covarianceFromConfig == true)) {
                            if (std::distance(mytok.begin(), mytok.end()) < size) {
                                if (rank == 0) throw std::runtime_error(("ERROR: Correlation matrix is of wrong size in Correlated Gaussian Observables: " + name + +" at line number:" + boost::lexical_cast<std::string>(lineNo) + ".\n").c_str());
                                else sleep(2);
                            }
                            if (lines.at(j)) {
                                myCorr(ni, nj) = atof((*beg).c_str());
                                nj++;
                            }
                            beg++;
                        } else {
                            if (rank == 0) throw std::runtime_error("ERROR: invalid correlation matrix for " + name + ". Check element (" + boost::lexical_cast<std::string>(ni + 1) + "," + boost::lexical_cast<std::string>(nj + 1) + ") in line number " + boost::lexical_cast<std::string>(lineNo) + " in file " + infilename + ".\n");
                            else sleep(2);
                        }
                    }
                    ni++;
                }
            }
            if (!myCorr.IsSymmetric()) {
                if (rank == 0) throw std::runtime_error("ERROR: invalid correlation matrix for " + name + ". Correlation matrix is not symmetric.\n");
                else sleep(2);
            } else {
                TMatrixDSym mySCorr(nlines);
                for (int i = 0; i < nlines; i++) {
                    for (int j = 0; j <= i; j++) {
                        mySCorr(i, j) = myCorr(i, j);
                        mySCorr(j, i) = mySCorr(i, j); // Make sure TMatrixDsym element (j,i) is stored symmetrically
                    }
                }
                ComputeCov(mySCorr);
            }
        } else {
            InvCov.ResizeTo(size, size);
        }
    } else {
        if (rank == 0) std::cout << "\nWARNING: Correlated (Gaussian) Observable " << name.c_str() << " defined with less than two observables" << " in file " << infilename << ". The set is being marked as normal Observables." << std::endl;
        if (getObs().size() == 1) Observables.push_back(new Observable(getObs(0)));
        for (int i = 0; i < size; i++) {
            IsEOF = getline(ifile, line).eof();
            lineNo++;
        }
    }
    return lineNo;
}

◆ setCovarianceFromConfig()

void CorrelatedGaussianObservables::setCovarianceFromConfig ( bool setInvCov )

inline

A method to specify whether the inverse covariance is being set from the config file.

Parameters

[in] setInvCov a boolean which is true if the inverse covariance matrix is set from file

Definition at line 176 of file CorrelatedGaussianObservables.h.

    {
        covarianceFromConfig = setInvCov;
    }

◆ setIsPrediction()

void CorrelatedGaussianObservables::setIsPrediction ( bool IsPrediction_i )

inline

A method to set a set of CGO to be predicted.

Parameters

[in] IsPrediction_i a boolean which is true if the set of CGO is set for prediction

Definition at line 149 of file CorrelatedGaussianObservables.h.

    {
        IsPrediction = IsPrediction_i;
    }

◆ setName()

void CorrelatedGaussianObservables::setName ( std::string name_i )

inline

A method to set a set of CGO to be predicted.

Parameters

[in] IsPrediction_i a boolean which is true if the set of CGO is set for prediction

Definition at line 167 of file CorrelatedGaussianObservables.h.

    {
        name = name_i;
    }

Member Data Documentation

◆ covarianceFromConfig

bool CorrelatedGaussianObservables::covarianceFromConfig

private

Definition at line 189 of file CorrelatedGaussianObservables.h.

◆ filepath

std::string CorrelatedGaussianObservables::filepath

private

The path to the config file being parsed.

Definition at line 186 of file CorrelatedGaussianObservables.h.

◆ InvCov

TMatrixDSym CorrelatedGaussianObservables::InvCov

private

The inverse covariance matrix.

Definition at line 183 of file CorrelatedGaussianObservables.h.

◆ IsEOF

bool CorrelatedGaussianObservables::IsEOF

private

A boolean which is true if the end of file is reached.

Definition at line 187 of file CorrelatedGaussianObservables.h.

◆ IsPrediction

bool CorrelatedGaussianObservables::IsPrediction

private

Flag to define a set of Correlated Observables to be predicted.

Definition at line 188 of file CorrelatedGaussianObservables.h.

◆ name

std::string CorrelatedGaussianObservables::name

private

The name of the correlated Gaussian Observables set.

Definition at line 185 of file CorrelatedGaussianObservables.h.

◆ Obs

std::vector<Observable> CorrelatedGaussianObservables::Obs

private

A vector of observables whose correlation will be calculated.

Definition at line 182 of file CorrelatedGaussianObservables.h.

◆ x

TVectorD CorrelatedGaussianObservables::x

private

The vector of the differences between the theory and average values of the observables in the set.

Definition at line 184 of file CorrelatedGaussianObservables.h.

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

Detailed Description

Public Member Functions

Private Attributes

Constructor & Destructor Documentation

◆ CorrelatedGaussianObservables() [1/3]

◆ CorrelatedGaussianObservables() [2/3]

◆ CorrelatedGaussianObservables() [3/3]

◆ ~CorrelatedGaussianObservables()

Member Function Documentation

◆ AddObs()

◆ ComputeCov()

◆ computeLeaveOneOutWeights()

◆ computeWeight()

◆ getCov()

◆ getName()

◆ getObs() [1/2]

◆ getObs() [2/2]

◆ isEOF()

◆ isPrediction()

◆ ParseCGO()

◆ setCovarianceFromConfig()

◆ setIsPrediction()

◆ setName()

Member Data Documentation

◆ covarianceFromConfig

◆ filepath

◆ InvCov

◆ IsEOF

◆ IsPrediction

◆ name

◆ Obs

◆ x