Read-only view of a subset of the entries of a GaussianParams or Gaussian. More...

#include <EmrysFilter/Distributions/Gaussian.h>

Public Types
using	Scalar = typename Source::Scalar
	the scalar type
using	Index = Eigen::Index
	Eigen's index type.
using	SpaceType = SubSpace
	the selected block's space
using	ParamsType = GaussianParams<SubSpace, Scalar>
	what copy() returns
using	MomentsType = typename ParamsType::MomentsType
	the mean and covariance together
using	VectorType = typename ParamsType::VectorType
	a value, such as the mean
using	MatrixType = typename ParamsType::MatrixType
	a matrix over differences
using	ResolvedType = typename internal::SelectionHolder<Source, Selection, SubSpace>::ResolvedType
	the selection, resolved
using	IndexVectorType = decltype(ResolvedType::indices)
	The selected mean entries' indices.
using	DifferenceIndexVectorType = decltype(ResolvedType::differenceIndices)
	The selected covariance rows' indices.

Public Member Functions
	GaussianMarginal (Source const &source, Selection const &selection, SubSpace space)
	Selects selection's coordinates, which form space.
	GaussianMarginal (Source const &source, Selection shared)
	Selects the coordinates of source that shared resolved, borrowing them.
Index	valueDimension () const
Index	differenceDimension () const
SubSpace const &	space () const
IndexVectorType const &	indices () const
DifferenceIndexVectorType const &	differenceIndices () const
auto	mean () const
auto	covarianceLower () const
MomentsType	moments () const
ParamsType	copy () const
std::expected< MatrixType, GaussianError >	covariance () const
std::expected< MatrixType, GaussianError >	precision () const
std::expected< MatrixType, GaussianError >	covarianceRoot () const
std::expected< MatrixType, GaussianError >	precisionRoot () const

Static Public Attributes
static constexpr int	ValueDimensionAtCompileTime = valueDimensionAtCompileTime<SubSpace>
	the length of a value, or Eigen::Dynamic
static constexpr int	MaxValueDimensionAtCompileTime = maxValueDimensionAtCompileTime<SubSpace>
	a bound on the length of a value, or Eigen::Dynamic
static constexpr int	DifferenceDimensionAtCompileTime = valueDimensionAtCompileTime<DifferenceSpace<SubSpace>>
	the length of a difference and the rows of the covariance, or Eigen::Dynamic
static constexpr int	MaxDifferenceDimensionAtCompileTime
	a bound on the length of a difference, or Eigen::Dynamic
static constexpr bool	IsInfoForm = false
	a block of the covariance, not of the precision [gaussian-forms]
static constexpr bool	IsSqrtForm = false
	a block of the covariance whole, not of a factor [gaussian-forms]

Detailed Description

template<typename Source, typename Selection, typename SubSpace>
class Emrys::GaussianMarginal< Source, Selection, SubSpace >

Read-only view of a subset of the entries of a GaussianParams or Gaussian.

Obtained from GaussianParams::marginal() or Gaussian::marginal(), and held by auto: its Selection is internal, so its type is never spelled, and a function that takes one takes it as a template parameter. Of GaussianParams it checks their shape (values may change underneath it, size may not); of a Gaussian it holds a copy of the view, and so the view's check. A marginal of a mixture or a hybrid gives one per component, which borrows the selection that marginal resolved once, and must not outlive it.

mean() and covarianceLower() return lazy Eigen expressions that must not outlive the marginal. covarianceLower() reads only the source's lower triangle. The dimensions are known at compile time when the selected block's sizes are. It is a GaussianForm, with the four fallible accessors of every form, so code generic over the forms takes it as it takes the other forms' marginals, which are parameters (decisions.md [form-densities]).

Template Parameters

Source	GaussianParams or Gaussian
Selection	an internal::Coordinates of the index sets of the selected block, from idx() for the mean and diffIdx() for the covariance, or an internal::SharedSelection of a selection resolved already
SubSpace	space of the selected block