Given a statistical model (prior) and a set of point constraints (likelihood), generate a new PCA model (posterior). More...

#include <PosteriorModelBuilder.h>

Inheritance diagram for statismo::PosteriorModelBuilder< T >:

Public Types
typedef Representer< T >	RepresenterType

typedef ModelBuilder< T >	Superclass

typedef Superclass::DataManagerType	DataManagerType

typedef Superclass::StatisticalModelType	StatisticalModelType

typedef RepresenterType::ValueType	ValueType

typedef RepresenterType::PointType	PointType

typedef StatisticalModelType::PointValueListType	PointValueListType

typedef DataManagerType::DataItemListType	DataItemListType

typedef MatrixType	PointCovarianceMatrixType

typedef StatisticalModelType::PointValuePairType	PointValuePairType

typedef std::pair < PointValuePairType, PointCovarianceMatrixType >	PointValueWithCovariancePairType

typedef std::list < PointValueWithCovariancePairType >	PointValueWithCovarianceListType

Public Types inherited from statismo::ModelBuilder< T >
typedef Representer< T >	RepresenterType

typedef StatisticalModel< T >	StatisticalModelType

typedef DataManager< T >	DataManagerType

typedef DataManagerType::DataItemListType	DataItemListType

Public Member Functions
void	Delete ()

virtual	~PosteriorModelBuilder ()

StatisticalModelType *	BuildNewModel (const DataItemListType &dataItemList, const PointValueListType &pointValues, double pointValueNoiseVariance, double noiseVariance) const

StatisticalModelType *	BuildNewModel (const DataItemListType &DataItemList, const PointValueWithCovarianceListType &pointValuesWithCovariance, double noiseVariance) const

StatisticalModelType *	BuildNewModelFromModel (const StatisticalModelType *model, const PointValueListType &pointValues, double pointValueNoiseVariance, bool computeScores=true) const

StatisticalModelType *	BuildNewModelFromModel (const StatisticalModelType *model, const PointValueWithCovarianceListType &pointValuesWithCovariance, bool computeScores=true) const

PointValueWithCovarianceListType	TrivialPointValueWithCovarianceListWithUniformNoise (const PointValueListType &pointValues, double pointValueNoiseVariance) const

Static Public Member Functions
static PosteriorModelBuilder *	Create ()

Additional Inherited Members
Static Public Attributes inherited from statismo::ModelBuilder< T >
static const double	TOLERANCE = 1e-5

Protected Member Functions inherited from statismo::ModelBuilder< T >
MatrixType	ComputeScores (const MatrixType &X, const StatisticalModelType *model) const

ModelInfo	CollectModelInfo () const

Detailed Description

template<typename T>
class statismo::PosteriorModelBuilder< T >

Given a statistical model (prior) and a set of point constraints (likelihood), generate a new PCA model (posterior).

This class builds a StatisticalModel, just as PCAModelBuilder. However, in addition to the data, this model builder also takes as input a set of point constraints, i.e. known values for points. The resulting model will satisfy these constraints, and thus has a much lower variability than an unconstrained model would have.

For mathematical detailes see the paper Posterior Shape Models Thomas Albrecht, Marcel L ̈thi, Thomas Gerig, Thomas Vetter Medical Image Analysis 2013

Add method that allows for the use of the pointId in the constraint.

Constructor & Destructor Documentation

template<typename T>

virtual statismo::PosteriorModelBuilder< T >::~PosteriorModelBuilder ( )

inlinevirtual

destructor

Member Function Documentation

template<typename T >

PosteriorModelBuilder< T >::StatisticalModelType * statismo::PosteriorModelBuilder< T >::BuildNewModel	(	const DataItemListType &	dataItemList,
		const PointValueListType &	pointValues,
		double	pointValueNoiseVariance,
		double	noiseVariance
	)		const

Builds a new model from the data provided in the dataManager, and the given constraints. This version of the function assumes a noise with a uniform uncorrelated variance of the form pointValueNoiseVariance * identityMatrix at every given point.

Parameters

DataItemList	The list holding the data the model is built from
pointValues	A list of (point, value) pairs with the known values.
pointValueNoiseVariance	The variance of the estimated error at the known points (the pointValues)
noiseVariance	The variance of the noise assumed on our data

Returns: a new statistical model

Warning: The returned model needs to be explicitly deleted by the user of this method.

template<typename T >

PosteriorModelBuilder< T >::StatisticalModelType * statismo::PosteriorModelBuilder< T >::BuildNewModel	(	const DataItemListType &	DataItemList,
		const PointValueWithCovarianceListType &	pointValuesWithCovariance,
		double	noiseVariance
	)		const

Builds a new model from the data provided in the dataManager, and the given constraints. For this version of the function, the covariance matrix of the noise needs to be specified for every point. These covariance matrices are passed in the pointValuesWithCovariance list.

Parameters

DataItemList	The list holding the data the model is built from
pointValuesWithCovariance	A list of ((point,value), covarianceMatrix) for each known value.
noiseVariance	The variance of the noise assumed on our data

Returns: a new statistical model

Warning: The returned model needs to be explicitly deleted by the user of this method.

template<typename T >

PosteriorModelBuilder< T >::StatisticalModelType * statismo::PosteriorModelBuilder< T >::BuildNewModelFromModel	(	const StatisticalModelType *	model,
		const PointValueListType &	pointValues,
		double	pointValueNoiseVariance,
		bool	computeScores = `true`
	)		const

Builds a new StatisticalModel given a StatisticalModel and the given constraints. If we interpret the given model as a prior distribution over the modeled objects, the resulting model can (loosely) be interpreted as the posterior distribution, after having observed the data given in the PointValues. This version of the function assumes a noise with a uniform uncorrelated variance of the form pointValueNoiseVariance * identityMatrix at every given point.

Parameters

model	A statistical model.
pointValues	A list of (point, value) pairs with the known values.
pointValueNoiseVariance	The variance of the estimated error at the known points (the pointValues)
computeScores	Determines whether the scores are computed and stored in the model.

Returns: a new statistical model

Warning: The returned model needs to be explicitly deleted by the user of this method.

template<typename T >

PosteriorModelBuilder< T >::StatisticalModelType * statismo::PosteriorModelBuilder< T >::BuildNewModelFromModel	(	const StatisticalModelType *	model,
		const PointValueWithCovarianceListType &	pointValuesWithCovariance,
		bool	computeScores = `true`
	)		const

Builds a new StatisticalModel given a StatisticalModel and the given constraints. If we interpret the given model as a prior distribution over the modeled objects, the resulting model can (loosely) be interpreted as the posterior distribution, after having observed the data given in the PointValues. For this version of the function, the covariance matrix of the noise needs to be specified for every point. These covariance matrices are passed in the pointValuesWithCovariance list.

Parameters

model	A statistical model.
pointValuesWithCovariance	A list of ((point,value), covarianceMatrix) for each known value.
computeScores	Determines whether the scores are computed and stored in the model.

Returns: a new statistical model

Warning: The returned model needs to be explicitly deleted by the user of this method.

template<typename T>

static PosteriorModelBuilder* statismo::PosteriorModelBuilder< T >::Create ( )

inlinestatic

Factory method to create a new PosteriorModelBuilder

Parameters

representer The representer

template<typename T>

void statismo::PosteriorModelBuilder< T >::Delete ( )

inline

Destroy the object. The same effect can be achieved by deleting the object in the usual way using the c++ delete keyword.

template<typename T >

PosteriorModelBuilder< T >::PointValueWithCovarianceListType statismo::PosteriorModelBuilder< T >::TrivialPointValueWithCovarianceListWithUniformNoise	(	const PointValueListType &	pointValues,
		double	pointValueNoiseVariance
	)		const

A convenience function to create a PointValueWithCovarianceList with uniform variance

Parameters

pointValues	A list of (point, value) pairs with the known values.
pointValueNoiseVariance	The variance of the estimated error at the known points (the pointValues)

Returns: a PointValueWithCovarianceListType with the given uniform variance

Warning: The returned model needs to be explicitly deleted by the user of this method.

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

Public Types

Public Member Functions

Static Public Member Functions

Additional Inherited Members

Detailed Description

template<typename T> class statismo::PosteriorModelBuilder< T >

Constructor & Destructor Documentation

Member Function Documentation

template<typename T>
class statismo::PosteriorModelBuilder< T >