cmp::distribution::NormalInverseWishartDistribution Class Reference

#include <distribution.h>

Inheritance diagram for cmp::distribution::NormalInverseWishartDistribution:

Collaboration diagram for cmp::distribution::NormalInverseWishartDistribution:

Public Member Functions
	NormalInverseWishartDistribution (const Eigen::Ref< const Eigen::VectorXd > &mean, double kappa, double nu, const Eigen::Ref< const Eigen::MatrixXd > &psi)
	NormalInverseWishartDistribution ()=default
double	logPDF (const Eigen::Ref< const Eigen::VectorXd > &x) const
Eigen::VectorXd	sample (std::default_random_engine &rng)
const Eigen::VectorXd &	mean () const
const double &	kappa () const
const double &	nu () const
const Eigen::MatrixXd	covariance () const
Eigen::MatrixXd	toCanonical (const Eigen::MatrixXd &x) const
Eigen::MatrixXd	toPhysical (const Eigen::MatrixXd &z) const
size_t	dimension () const
Public Member Functions inherited from cmp::distribution::MultivariateDistribution< NormalInverseWishartDistribution >
double	logPDF (const Eigen::Ref< const Eigen::VectorXd > &x) const
	Computes the joint log probability density function (log-PDF) of the distribution.
Eigen::VectorXd	sample (std::default_random_engine &rng)
	Draws a single vector sample from the joint distribution.
Eigen::MatrixXd	toCanonical (const Eigen::MatrixXd &x) const
	Transforms physical samples to standard canonical space.
Eigen::MatrixXd	fromCanonical (const Eigen::MatrixXd &x) const
	Transforms standard canonical samples back to physical space.
size_t	dimension () const
	Returns the dimensionality of the multivariate space.

Static Public Member Functions
static NormalInverseWishartDistribution	canonical (const size_t dim, double kappa, double nu)
static NormalInverseWishartDistribution	empiricalPrior (const Eigen::MatrixXd &data, double expected_clusters, double kappa0)
	Computes an empirical Normal-Inverse-Wishart prior from the provided dataset. Useful for initializing the prior in a Dirichlet Process Mixture Model (DPMM) when you have observed data.

Private Attributes
Eigen::VectorXd	mean_
	Mean parameter vector.
double	kappa_
	Degrees of freedom scaling parameter kappa.
double	nu_
	Degrees of freedom parameter nu.
size_t	dim_
	Dimensionality of the parameter space.
Eigen::LDLT< Eigen::MatrixXd >	covLDLT_
	LDLT decomposition of the scaling matrix.
double	logDeterminant_
	Log-determinant of the scale covariance matrix.
std::normal_distribution< double >	distN_ {0., 1.}
	Normal distribution helper for coordinate sampling.

Constructor & Destructor Documentation

NormalInverseWishartDistribution() [1/2]

cmp::distribution::NormalInverseWishartDistribution::NormalInverseWishartDistribution ( const Eigen::Ref< const Eigen::VectorXd > &  mean, double  kappa, double  nu, const Eigen::Ref< const Eigen::MatrixXd > &  psi  )

inline

NormalInverseWishartDistribution() [2/2]

cmp::distribution::NormalInverseWishartDistribution::NormalInverseWishartDistribution ( )

default

Member Function Documentation

canonical()

static NormalInverseWishartDistribution cmp::distribution::NormalInverseWishartDistribution::canonical ( const size_t  dim, double  kappa, double  nu  )

inlinestatic

covariance()

const Eigen::MatrixXd cmp::distribution::NormalInverseWishartDistribution::covariance ( ) const

inline

dimension()

size_t cmp::distribution::NormalInverseWishartDistribution::dimension ( ) const

inline

empiricalPrior()

static NormalInverseWishartDistribution cmp::distribution::NormalInverseWishartDistribution::empiricalPrior ( const Eigen::MatrixXd &  data, double  expected_clusters, double  kappa0  )

inlinestatic

Computes an empirical Normal-Inverse-Wishart prior from the provided dataset. Useful for initializing the prior in a Dirichlet Process Mixture Model (DPMM) when you have observed data.

Parameters

data	The observed data matrix.
expected_clusters	The expected number of clusters.
kappa0	The prior precision parameter.

Returns

The computed empirical Normal-Inverse-Wishart prior.

kappa()

const double & cmp::distribution::NormalInverseWishartDistribution::kappa ( ) const

inline

logPDF()

double cmp::distribution::NormalInverseWishartDistribution::logPDF ( const Eigen::Ref< const Eigen::VectorXd > &  x ) const

inline

mean()

const Eigen::VectorXd & cmp::distribution::NormalInverseWishartDistribution::mean ( ) const

inline

nu()

const double & cmp::distribution::NormalInverseWishartDistribution::nu ( ) const

inline

sample()

Eigen::VectorXd cmp::distribution::NormalInverseWishartDistribution::sample ( std::default_random_engine &  rng )

inline

toCanonical()

Eigen::MatrixXd cmp::distribution::NormalInverseWishartDistribution::toCanonical ( const Eigen::MatrixXd &  x ) const

inline

toPhysical()

Eigen::MatrixXd cmp::distribution::NormalInverseWishartDistribution::toPhysical ( const Eigen::MatrixXd &  z ) const

inline

Member Data Documentation

covLDLT_

Eigen::LDLT<Eigen::MatrixXd> cmp::distribution::NormalInverseWishartDistribution::covLDLT_

private

LDLT decomposition of the scaling matrix.

dim_

size_t cmp::distribution::NormalInverseWishartDistribution::dim_

private

Dimensionality of the parameter space.

distN_

std::normal_distribution<double> cmp::distribution::NormalInverseWishartDistribution::distN_ {0., 1.}

private

Normal distribution helper for coordinate sampling.

kappa_

double cmp::distribution::NormalInverseWishartDistribution::kappa_

private

Degrees of freedom scaling parameter kappa.

logDeterminant_

double cmp::distribution::NormalInverseWishartDistribution::logDeterminant_

private

Log-determinant of the scale covariance matrix.

mean_

Eigen::VectorXd cmp::distribution::NormalInverseWishartDistribution::mean_

private

Mean parameter vector.

nu_

double cmp::distribution::NormalInverseWishartDistribution::nu_

private

Degrees of freedom parameter nu.

---

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

• include/distribution.h