A class for handling scalar Gaussian random fields (GRF). More...

#include <ScalarGaussianRandomField.h>

Collaboration diagram for QUESO::ScalarGaussianRandomField< V, M >:

Public Member Functions
Constructor/Destructor methods
	ScalarGaussianRandomField (const char *prefix, const VectorSet< V, M > &indexSet, const BaseScalarFunction< V, M > &meanFunction, const BaseScalarCovarianceFunction< V, M > &covarianceFunction)
	Constructor. More...

	ScalarGaussianRandomField (const ScalarGaussianRandomField &obj)
	TODO: Copy constructor. More...

	~ScalarGaussianRandomField ()
	Destructor. More...

Set methods
ScalarGaussianRandomField &	operator= (const ScalarGaussianRandomField &rhs)
	TODO: Assignment operator; it copies `rhs` to `this`. More...

Math methods
const VectorSet< V, M > &	indexSet () const
	Index set; access to protected attribute m_indexSet. More...

const BaseScalarFunction< V, M > &	meanFunction () const
	Mean function; access to protected attribute m_meanFunction. More...

const BaseScalarCovarianceFunction < V, M > &	covarianceFunction () const
	Covariance function; access to protected attribute m_covarianceFunction. More...

void	sampleFunction (const std::vector< V * > &fieldPositions, V &sampleValues)
	Function that samples from a Gaussian PDF. More...

Protected Member Functions
void	copy (const ScalarGaussianRandomField &src)
	Copy method. More...

Protected Attributes
const BaseEnvironment &	m_env
	Environment. More...

std::string	m_prefix
	Prefix. More...

const VectorSet< V, M > &	m_indexSet
	Index set. More...

const BaseScalarFunction< V, M > &	m_meanFunction
	Mean function. More...

const BaseScalarCovarianceFunction < V, M > &	m_covarianceFunction
	Covariance function. More...

std::vector< V * >	m_savedPositions
	Saved positions. More...

VectorSpace< V, M > *	m_savedRvImageSpace
	Image set of the RV. More...

V *	m_savedRvLawExpVector
	Vector of the mean value of the RV. More...

M *	m_savedRvLawCovMatrix
	Covariance matrix of the RV. More...

GaussianVectorRV< V, M > *	m_savedRv
	My RV. More...

Detailed Description

template<class V, class M>
class QUESO::ScalarGaussianRandomField< V, M >

A class for handling scalar Gaussian random fields (GRF).

This class implements a scalar Gaussian random field (GRF); i.e. a random field involving Gaussian probability density functions (PDFs) of the variables. A one-dimensional GRF is also called a Gaussian process.

Definition at line 46 of file ScalarGaussianRandomField.h.

Constructor & Destructor Documentation

template<class V , class M >

QUESO::ScalarGaussianRandomField< V, M >::ScalarGaussianRandomField	(	const char *	prefix,
		const VectorSet< V, M > &	indexSet,
		const BaseScalarFunction< V, M > &	meanFunction,
		const BaseScalarCovarianceFunction< V, M > &	covarianceFunction
	)

Constructor.

Constructs a new object, given a prefix, an index set, and both a mean and a covariance function. This method deletes the previous saved positions.

Definition at line 32 of file ScalarGaussianRandomField.C.

References QUESO::ScalarGaussianRandomField< V, M >::m_savedPositions.

   :
   m_env                (indexSet.env()),
   m_prefix             ((std::string)(prefix)+"grf_"),
   m_indexSet           (indexSet),
   m_meanFunction       (meanFunction),
   m_covarianceFunction (covarianceFunction),
   m_savedRvImageSpace  (NULL),
   m_savedRvLawExpVector(NULL),
   m_savedRvLawCovMatrix(NULL),
   m_savedRv            (NULL)
 {
   m_savedPositions.clear();
 }

template<class V , class M >

QUESO::ScalarGaussianRandomField< V, M >::ScalarGaussianRandomField ( const ScalarGaussianRandomField< V, M > & obj )

TODO: Copy constructor.

Todo:: : implement me!

template<class V , class M >

QUESO::ScalarGaussianRandomField< V, M >::~ScalarGaussianRandomField ( )

Destructor.

Definition at line 52 of file ScalarGaussianRandomField.C.

53 {

54 }

Member Function Documentation

template<class V , class M >

void QUESO::ScalarGaussianRandomField< V, M >::copy ( const ScalarGaussianRandomField< V, M > & src )

protected

Copy method.

template<class V , class M >

const BaseScalarCovarianceFunction< V, M > & QUESO::ScalarGaussianRandomField< V, M >::covarianceFunction ( ) const

Covariance function; access to protected attribute m_covarianceFunction.

Definition at line 72 of file ScalarGaussianRandomField.C.

 {
   return m_covarianceFunction;
 }

template<class V , class M >

const VectorSet< V, M > & QUESO::ScalarGaussianRandomField< V, M >::indexSet ( ) const

Index set; access to protected attribute m_indexSet.

Definition at line 58 of file ScalarGaussianRandomField.C.

 {
   return m_indexSet;
 }

template<class V , class M >

const BaseScalarFunction< V, M > & QUESO::ScalarGaussianRandomField< V, M >::meanFunction ( ) const

Mean function; access to protected attribute m_meanFunction.

Definition at line 65 of file ScalarGaussianRandomField.C.

 {
   return m_meanFunction;
 }

template<class V , class M >

ScalarGaussianRandomField& QUESO::ScalarGaussianRandomField< V, M >::operator= ( const ScalarGaussianRandomField< V, M > & rhs )

TODO: Assignment operator; it copies rhs to this.

Todo:: : implement me!

template<class V , class M >

void QUESO::ScalarGaussianRandomField< V, M >::sampleFunction	(	const std::vector< V * > &	fieldPositions,
		V &	sampleValues
	)

Function that samples from a Gaussian PDF.

Given the field positions, this method performs a number of tests, calculates the mean vector, the covariance matrix and then it samples from a Gaussian random vector as many positions as required.

Definition at line 79 of file ScalarGaussianRandomField.C.

References UQ_FATAL_TEST_MACRO.

 {
   if ((m_env.subDisplayFile()) && (m_env.displayVerbosity() >= 99)) {
     *m_env.subDisplayFile() << "Entering ScalarGaussianRandomField<V,M>::sampleFunction()"
                             << std::endl;
   }
 
   UQ_FATAL_TEST_MACRO(fieldPositions.size() != sampleValues.sizeLocal(),
                       m_env.fullRank(),
                       "ScalarGaussianRandomField<V,M>::sampleFunction()",
                       "invalid input data");
 
   if ((m_savedPositions.size() == 0   ) &&
       (m_savedRvImageSpace     == NULL) &&
       (m_savedRvLawExpVector   == NULL) &&
       (m_savedRvLawCovMatrix   == NULL) &&
       (m_savedRv               == NULL)) {
     // Ok
   }
   else if ((m_savedPositions.size() != 0   ) &&
            (m_savedRvImageSpace     != NULL) &&
            (m_savedRvLawExpVector   != NULL) &&
            (m_savedRvLawCovMatrix   != NULL) &&
            (m_savedRv               != NULL)) {
     // Ok
   }
   else {
     UQ_FATAL_TEST_MACRO(true,
                         m_env.fullRank(),
                         "ScalarGaussianRandomField<V,M>::sampleFunction()",
                         "invalid combination of pointer values");
   }
 
   unsigned int numberOfPositions = fieldPositions.size();
   bool instantiate = true;
   if (m_savedPositions.size() == numberOfPositions) {
     bool allPositionsAreEqual = true;
     for (unsigned int i = 0; i < numberOfPositions; ++i) {
       UQ_FATAL_TEST_MACRO(m_savedPositions[i] == NULL,
                           m_env.fullRank(),
                           "ScalarGaussianRandomField<V,M>::sampleFunction()",
                           "m_savedPositions[i] should not be NULL");
       if ((m_savedPositions[i]->sizeLocal() == fieldPositions[i]->sizeLocal()) &&
           (*(m_savedPositions[i])           == *(fieldPositions[i])          )) {
         // Ok
       }
       else {
         allPositionsAreEqual = false;
         break;
       }
     } // for i
     instantiate = !allPositionsAreEqual;
   }
 
   if (instantiate) {
     delete m_savedRv;
     delete m_savedRvLawCovMatrix;
     delete m_savedRvLawExpVector;
     delete m_savedRvImageSpace;
     for (unsigned int i = 0; i < m_savedPositions.size(); ++i) {
       delete m_savedPositions[i];
     }
     m_savedPositions.clear();
 
     // Set m_savedPositions
     m_savedPositions.resize(numberOfPositions,NULL);
     for (unsigned int i = 0; i < m_savedPositions.size(); ++i) {
       m_savedPositions[i] = new V(*(fieldPositions[i]));
     }
 
     // Set m_savedRvImageSpace
     m_savedRvImageSpace = new VectorSpace<V,M>(m_env, "grf_", numberOfPositions, NULL);
 
     // Set m_savedRvLawExpVector
     m_savedRvLawExpVector = new V(m_savedRvImageSpace->zeroVector());
     for (unsigned int i = 0; i < numberOfPositions; ++i) {
       (*m_savedRvLawExpVector)[i] = m_meanFunction.actualValue(*(fieldPositions[i]),NULL,NULL,NULL,NULL);
     }
 
     // Set m_savedRvLawCovMatrix
     m_savedRvLawCovMatrix = new M(m_savedRvImageSpace->zeroVector());
     for (unsigned int i = 0; i < numberOfPositions; ++i) {
       for (unsigned int j = 0; j < numberOfPositions; ++j) {
         (*m_savedRvLawCovMatrix)(i,j) = m_covarianceFunction.value(*(fieldPositions[i]),*(fieldPositions[j]));
         if ((m_env.subDisplayFile()) && (m_env.displayVerbosity() >= 3)) {
           *m_env.subDisplayFile() << "In ScalarGaussianRandomField<V,M>::sampleFunction()"
                                   << ": i = " << i
                                   << ", j = " << j
                                   << ", *(fieldPositions[i]) = " << *(fieldPositions[i])
                                   << ", *(fieldPositions[j]) = " << *(fieldPositions[j])
                                   << ", (*m_savedRvLawCovMatrix)(i,j) = " << (*m_savedRvLawCovMatrix)(i,j)
                                   << std::endl;
         }
       }
     }
 
     // Set m_savedRv
     m_savedRv = new GaussianVectorRV<V,M>("grf_",
                                                  *m_savedRvImageSpace,
                                                  *m_savedRvLawExpVector,
                                                  *m_savedRvLawCovMatrix);
 
     if ((m_env.subDisplayFile()) && (m_env.displayVerbosity() >= 3)) {
       *m_env.subDisplayFile() << "In ScalarGaussianRandomField<V,M>::sampleFunction()"
                               << ": just instantiated Gaussian RV"
                               << "\n *m_savedRvLawExpVector = " << *m_savedRvLawExpVector
                               << "\n *m_savedRvLawCovMatrix = " << *m_savedRvLawCovMatrix
                               << std::endl;
       for (unsigned int i = 0; i < numberOfPositions; ++i) {
         *m_env.subDisplayFile() << " *(m_savedPositions[" << i
                                 << "]) = "                << *(m_savedPositions[i])
                                 << std::endl;
       }
     }
   } // if (instantiate)
 
   // Generate sample function
   if ((m_env.subDisplayFile()) && (m_env.displayVerbosity() >= 99)) {
     *m_env.subDisplayFile() << "In ScalarGaussianRandomField<V,M>::sampleFunction()"
                             << ": about to realize sample values"
                             << std::endl;
   }
   m_savedRv->realizer().realization(sampleValues);
   if ((m_env.subDisplayFile()) && (m_env.displayVerbosity() >= 99)) {
     *m_env.subDisplayFile() << "In ScalarGaussianRandomField<V,M>::sampleFunction()"
                             << ": just realized sample values"
                             << std::endl;
   }
 
   if ((m_env.subDisplayFile()) && (m_env.displayVerbosity() >= 99)) {
     *m_env.subDisplayFile() << "Leaving ScalarGaussianRandomField<V,M>::sampleFunction()"
                             << std::endl;
   }
 
   return;
 }