Class defining infinite dimensional Gaussian measures. More...

#include <InfiniteDimensionalGaussian.h>

Inheritance diagram for QUESO::InfiniteDimensionalGaussian:

Collaboration diagram for QUESO::InfiniteDimensionalGaussian:

Public Member Functions
virtual boost::shared_ptr < FunctionBase >	draw ()
	Draw from the measure, and then return a shared pointer to the draw. More...

virtual double	get_kl_coefficient (unsigned int i) const
	Return coefficient `i` of the KL expansion of the current draw. Must be called after draw() More...

Constructor/Destructor methods
	InfiniteDimensionalGaussian (const FullEnvironment &env, const FunctionBase &mean, const OperatorBase &precision, double alpha, double beta)
	Construct a Gaussian with mean `mean`, precision operator `precision`. `alpha` controls regularity of draws and `beta^2` is a multiplicative coefficient of `precision`. More...

	~InfiniteDimensionalGaussian ()
	Destructor. More...

Public Member Functions inherited from QUESO::InfiniteDimensionalMeasureBase
	InfiniteDimensionalMeasureBase ()
	Default constructor. More...

virtual	~InfiniteDimensionalMeasureBase ()
	Destructor. More...

Private Attributes
const FunctionBase &	mean

const OperatorBase &	precision

const FullEnvironment &	env

double	alpha

double	beta

std::vector< double >	coeffs

Detailed Description

Class defining infinite dimensional Gaussian measures.

Definition at line 50 of file InfiniteDimensionalGaussian.h.

Constructor & Destructor Documentation

QUESO::InfiniteDimensionalGaussian::InfiniteDimensionalGaussian	(	const FullEnvironment &	env,
		const FunctionBase &	mean,
		const OperatorBase &	precision,
		double	alpha,
		double	beta
	)

Construct a Gaussian with mean mean, precision operator precision. alpha controls regularity of draws and beta^2 is a multiplicative coefficient of precision.

The argument alpha refers to the power to which the precision will be taken. And beta is the multiplicative coefficient of preicision. The complete measure is therefore:

N(mean, beta^2 * pow(precision, alpha))

It is expected that mean and precision will live longer than this

Definition at line 38 of file InfiniteDimensionalGaussian.C.

References coeffs, and QUESO::OperatorBase::get_num_converged().

   : InfiniteDimensionalMeasureBase(),
     mean(mean),
     precision(precision),
     env(env),
     alpha(alpha),
     beta(beta)
 {
   this->coeffs.resize(this->precision.get_num_converged(), 0.0);
 }

QUESO::InfiniteDimensionalGaussian::~InfiniteDimensionalGaussian ( )

Destructor.

Definition at line 54 of file InfiniteDimensionalGaussian.C.

55 {

56 }

Member Function Documentation

boost::shared_ptr< FunctionBase > QUESO::InfiniteDimensionalGaussian::draw ( )

virtual

Draw from the measure, and then return a shared pointer to the draw.

Implements QUESO::InfiniteDimensionalMeasureBase.

Definition at line 58 of file InfiniteDimensionalGaussian.C.

References beta, coeffs, env, QUESO::RngBase::gaussianSample(), QUESO::OperatorBase::get_num_converged(), QUESO::OperatorBase::inverse_kl_transform(), precision, and QUESO::BaseEnvironment::rngObject().

 {
   unsigned int i;
 
   for (i = 0; i < this->precision.get_num_converged(); i++) {
     (this->coeffs)[i] = env.rngObject()->gaussianSample(this->beta);
   }
 
   boost::shared_ptr<FunctionBase> f(this->precision.inverse_kl_transform(this->coeffs, this->alpha));
   return f;
 }

double QUESO::InfiniteDimensionalGaussian::get_kl_coefficient ( unsigned int i ) const

virtual

Return coefficient i of the KL expansion of the current draw. Must be called after draw()

Implements QUESO::InfiniteDimensionalMeasureBase.

Definition at line 70 of file InfiniteDimensionalGaussian.C.

References alpha, coeffs, QUESO::OperatorBase::get_eigenvalue(), and precision.

 {
   // This is code repetition, but I'm not quite sure this belongs
   // in the operator class, because it's useful in the measure
   return (this->coeffs)[i] / pow(this->precision.get_eigenvalue(i), this->alpha / 2.0);
 }