Defines an abstract Gaussian class.@seealso Gaussian. More...

#include <GaussianBase.h>

Inheritance diagram for muq::Modeling::GaussianBase:

Detailed Description

Defines an abstract Gaussian class.

Definition at line 15 of file GaussianBase.h.

Public Member Functions
	GaussianBase (unsigned int dim)

	GaussianBase (unsigned int dim, Eigen::VectorXi const &hyperSizesIn)

	GaussianBase (Eigen::VectorXd const &meanIn)

	GaussianBase (Eigen::VectorXd const &meanIn, Eigen::VectorXi const &hyperSizesIn)

virtual	~GaussianBase ()=default

virtual unsigned int	Dimension () const

virtual Eigen::MatrixXd	ApplyCovariance (Eigen::Ref< const Eigen::MatrixXd > const &x) const =0

virtual Eigen::MatrixXd	ApplyPrecision (Eigen::Ref< const Eigen::MatrixXd > const &x) const =0

virtual Eigen::MatrixXd	ApplyCovSqrt (Eigen::Ref< const Eigen::MatrixXd > const &x) const =0

virtual Eigen::MatrixXd	ApplyPrecSqrt (Eigen::Ref< const Eigen::MatrixXd > const &x) const =0

virtual Eigen::VectorXd const &	GetMean () const

virtual void	SetMean (Eigen::VectorXd const &newMu)

virtual double	LogDeterminant () const

virtual void	ResetHyperparameters (ref_vector< Eigen::VectorXd > const &params)

virtual Eigen::VectorXd	GradLogDensityImpl (unsigned int wrt, ref_vector< Eigen::VectorXd > const &inputs) override

virtual Eigen::VectorXd	ApplyLogDensityHessianImpl (unsigned int wrt1, unsigned int wrt2, ref_vector< Eigen::VectorXd > const &inputs, Eigen::VectorXd const &vec) override

Public Member Functions inherited from muq::Modeling::Distribution
	Distribution (unsigned int varSizeIn, Eigen::VectorXi const &hyperSizesIn=Eigen::VectorXi())

virtual	~Distribution ()=default

virtual double	LogDensity (ref_vector< Eigen::VectorXd > const &inputs)
	Evaluate the log-density. More...

virtual double	LogDensity (std::vector< Eigen::VectorXd > const &inputs)

	VARIADIC_TO_REFVECTOR (LogDensity, Eigen::VectorXd, double)

virtual Eigen::VectorXd	GradLogDensity (unsigned int wrt, std::vector< Eigen::VectorXd > const &inputs)

virtual Eigen::VectorXd	GradLogDensity (unsigned int wrt, ref_vector< Eigen::VectorXd > const &inputs)

template<typename... Args>
Eigen::VectorXd	GradLogDensity (unsigned int wrt, Args... args)

virtual Eigen::VectorXd	ApplyLogDensityHessian (unsigned int const inWrt1, unsigned int const inWrt2, ref_vector< Eigen::VectorXd > const &input, Eigen::VectorXd const &vec)

virtual Eigen::VectorXd	ApplyLogDensityHessian (unsigned int const inWrt1, unsigned int const inWrt2, std::vector< Eigen::VectorXd > const &input, Eigen::VectorXd const &vec)

Eigen::VectorXd	Sample (ref_vector< Eigen::VectorXd > const &inputs)
	Sample the distribution. More...

Eigen::VectorXd	Sample (std::vector< Eigen::VectorXd > const &inputs)

Eigen::VectorXd	Sample ()
	Sample the distribution with no inputs. More...

	VARIADIC_TO_REFVECTOR (Sample, Eigen::VectorXd, Eigen::VectorXd)

std::shared_ptr< Density >	AsDensity ()
	Returns a density built from this distribution. More...

std::shared_ptr< RandomVariable >	AsVariable ()
	Returns a random variable built from this distribution. More...

Additional Inherited Members
Public Attributes inherited from muq::Modeling::Distribution
const unsigned int	varSize

const Eigen::VectorXi	hyperSizes

Constructor & Destructor Documentation

◆ GaussianBase() [1/4]

GaussianBase::GaussianBase ( unsigned int dim )

Basic constructor for Gaussians with not additional hyperparameters.

Parameters

[in] dim The dimension of the Gaussian distribution.

Definition at line 8 of file GaussianBase.cpp.

◆ GaussianBase() [2/4]

GaussianBase::GaussianBase	(	unsigned int	dim,
		Eigen::VectorXi const &	hyperSizesIn
	)

Basic constructor for Gaussians with hyperparameter inputs.

Parameters

[in]	dim	The dimension of the Gaussian distribution
[in]	hyperSizesIn	A vector of integers specify the size of any additional inputs (e.g., mean, standard deviation).

Definition at line 12 of file GaussianBase.cpp.

◆ GaussianBase() [3/4]

GaussianBase::GaussianBase ( Eigen::VectorXd const & meanIn )

Construct a Gaussian with no hyperparameter inputs and a specified mean vector.

Parameters

[in] meanIn A vector containing the mean of the Gaussian distribution.

Definition at line 19 of file GaussianBase.cpp.

◆ GaussianBase() [4/4]

GaussianBase::GaussianBase	(	Eigen::VectorXd const &	meanIn,
		Eigen::VectorXi const &	hyperSizesIn
	)

Construct a Gaussian with hyperparameters and a specified mean vector.

Parameters

[in]	meanIn	A vector containing the mean of the Gaussian distribution.
[in]	hyperSizesIn	A vector of integers specify the size of any additional inputs (e.g., mean, standard deviation).

Definition at line 24 of file GaussianBase.cpp.

◆ ~GaussianBase()

virtual muq::Modeling::GaussianBase::~GaussianBase ( )

virtualdefault

Member Function Documentation

◆ ApplyCovariance()

virtual Eigen::MatrixXd muq::Modeling::GaussianBase::ApplyCovariance ( Eigen::Ref< const Eigen::MatrixXd > const & x ) const

pure virtual

Applies the covariance matrix to a vector \(x\)

Parameters

[in] x A reference to the input vector

Returns: \(\Sigma x\), where \(\Sigma\) is the covariance matrix.

Implemented in muq::Modeling::Gaussian.

◆ ApplyCovSqrt()

virtual Eigen::MatrixXd muq::Modeling::GaussianBase::ApplyCovSqrt ( Eigen::Ref< const Eigen::MatrixXd > const & x ) const

pure virtual

Applies a matrix square root of the covariance to a vector x.

Parameters

[in] x A reference to the input vector

Returns: \(\Sigma^{1/2} x\), where \(\Sigma^{1/2}\) is a square root of the covariance matrix (e.g., Cholesky factor).

Implemented in muq::Modeling::Gaussian.

◆ ApplyLogDensityHessianImpl()

Eigen::VectorXd GaussianBase::ApplyLogDensityHessianImpl	(	unsigned int	wrt1,
		unsigned int	wrt2,
		ref_vector< Eigen::VectorXd > const &	inputs,
		Eigen::VectorXd const &	vec
	)

overridevirtual

Reimplemented from muq::Modeling::Distribution.

Definition at line 69 of file GaussianBase.cpp.

References ApplyPrecision().

◆ ApplyPrecision()

virtual Eigen::MatrixXd muq::Modeling::GaussianBase::ApplyPrecision ( Eigen::Ref< const Eigen::MatrixXd > const & x ) const

pure virtual

Applies the precision matrix (inverse covariance) to a vector \(x\)

Parameters

[in] x A reference to the input vector

Returns: \(\Sigma^{-1} x\), where \(\Sigma\) is the covariance matrix.

Implemented in muq::Modeling::Gaussian.

Referenced by ApplyLogDensityHessianImpl(), and GradLogDensityImpl().

◆ ApplyPrecSqrt()

virtual Eigen::MatrixXd muq::Modeling::GaussianBase::ApplyPrecSqrt ( Eigen::Ref< const Eigen::MatrixXd > const & x ) const

pure virtual

Applies a matrix square root of the precision to a vector x. If \(\Sigma= L L^T\) for a covariance matrix \(\Sigma\) and a square root \(L\), then \(\Sigma^{-1} = L^{-T}L^{-1}\) is a decomposition of the precision matrix and \(L^{-T}\) is a square root of the precision matrix. This function returns the action of \(L^{-T}\) on a vector \(x\).

Parameters

[in] x A reference to the input vector

Returns: \(L^{-T} x\)

Implemented in muq::Modeling::Gaussian.

◆ Dimension()

unsigned int GaussianBase::Dimension ( ) const

virtual

Returns: The dimension of the distribution.

Definition at line 48 of file GaussianBase.cpp.

References mean.

Referenced by muq::Modeling::Gaussian::Condition().

◆ GetMean()

virtual Eigen::VectorXd const& muq::Modeling::GaussianBase::GetMean ( ) const

inlinevirtual

Returns a vector containing the mean of the Gaussian.

Returns: A vector containing the mean.

Definition at line 91 of file GaussianBase.h.

References mean.

◆ GradLogDensityImpl()

Eigen::VectorXd GaussianBase::GradLogDensityImpl	(	unsigned int	wrt,
		ref_vector< Eigen::VectorXd > const &	inputs
	)

overridevirtual

Compute the gradient of the log density with respect to either the distribution input or the hyperparameters. This should be overridden by child classes that have hyperparameters. If not overridden, this function will assert false if wrt>0.

Parameters

[in] wrt Specifies the index of the variable we wish to take the gradient wrt. If wrt==0, then the gradient should be taken wrt the input variable.

Returns: The gradient of the log density.

Reimplemented from muq::Modeling::Distribution.

Definition at line 58 of file GaussianBase.cpp.

References ApplyPrecision(), and mean.

◆ LogDensityImpl()

double GaussianBase::LogDensityImpl ( ref_vector< Eigen::VectorXd > const & inputs )

overrideprotectedvirtual

Compute the log density.

Parameters

[inputs] A vector of extra hyperparameter vectors.

Returns: A double containing the log density.

Reimplemented from muq::Modeling::Distribution.

Definition at line 30 of file GaussianBase.cpp.

References mean, and ResetHyperparameters().

◆ LogDeterminant()

virtual double muq::Modeling::GaussianBase::LogDeterminant ( ) const

inlinevirtual

Return the log determinant of the covariance matrix. Defaults to 0.0, so this should be overridden by child classes that are used in applications where the density needs to known completely, not only up to a normalizing constant.

Reimplemented in muq::Modeling::Gaussian.

Definition at line 105 of file GaussianBase.h.

◆ ResetHyperparameters()

virtual void muq::Modeling::GaussianBase::ResetHyperparameters ( ref_vector< Eigen::VectorXd > const & params )

inlinevirtual

Process the hyperparameter inputs. This should be overridden by children that have extra inputs.

Parameters

[in] params A vector of hyperparameter vectors.

Reimplemented in muq::Modeling::Gaussian.

Definition at line 111 of file GaussianBase.h.

Referenced by LogDensityImpl(), and SampleImpl().

◆ SampleImpl()

Eigen::VectorXd GaussianBase::SampleImpl ( ref_vector< Eigen::VectorXd > const & inputs )

overrideprotectedvirtual

Sample the distribution.

Reimplemented from muq::Modeling::Distribution.

Definition at line 39 of file GaussianBase.cpp.

References mean, and ResetHyperparameters().

◆ SetMean()

void GaussianBase::SetMean ( Eigen::VectorXd const & newMu )

virtual

Set the mean of this Gaussian. The new mean must be the same size as the old mean.

Parameters

[in] newMu The new mean.

Definition at line 52 of file GaussianBase.cpp.

References mean.

Member Data Documentation

◆ mean

Eigen::VectorXd muq::Modeling::GaussianBase::mean

protected

Definition at line 140 of file GaussianBase.h.

Referenced by muq::Modeling::Gaussian::Condition(), Dimension(), muq::Modeling::Gaussian::Gaussian(), muq::Modeling::Gaussian::GetCovariance(), GetMean(), muq::Modeling::Gaussian::GetPrecision(), GradLogDensityImpl(), LogDensityImpl(), muq::Modeling::Gaussian::ResetHyperparameters(), SampleImpl(), muq::Modeling::Gaussian::SetCovariance(), SetMean(), and muq::Modeling::Gaussian::SetPrecision().

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

doxygen_prep/MUQ/Modeling/Distributions/GaussianBase.h
doxygen_prep/modules/Modeling/src/Distributions/GaussianBase.cpp

Detailed Description

Public Member Functions

Additional Inherited Members

Constructor & Destructor Documentation

◆ GaussianBase() [1/4]

◆ GaussianBase() [2/4]

◆ GaussianBase() [3/4]

◆ GaussianBase() [4/4]

◆ ~GaussianBase()

Member Function Documentation

◆ ApplyCovariance()

◆ ApplyCovSqrt()

◆ ApplyLogDensityHessianImpl()

◆ ApplyPrecision()

◆ ApplyPrecSqrt()

◆ Dimension()

◆ GetMean()

◆ GradLogDensityImpl()

◆ LogDensityImpl()

◆ LogDeterminant()

◆ ResetHyperparameters()

◆ SampleImpl()

◆ SetMean()

Member Data Documentation

◆ mean