Spectral clustering. More...

#include <CRNSpectralClustering.h>

Public Member Functions
	SpectralClustering (const SpectralClustering &)=delete

	SpectralClustering (SpectralClustering &&)=default

SpectralClustering &	operator= (const SpectralClustering &)=delete

SpectralClustering &	operator= (SpectralClustering &&)=default

const std::multimap< double, MatrixDouble > &	GetEigenpairs () const noexcept
	Gets the data projected on each coordinates (higher eigenvalues are associated the to most significant coordinates) More...

std::vector< double >	GetEigenvalues () const
	Returns the eigenvalues (sorted from highest to lowest) More...

size_t	EstimateClusterNumber (double limit=1.0) const
	Estimates the number of clusters. More...

std::vector< std::vector < double > >	ProjectData (size_t ncoordinates, bool normalize) const
	Projects the data on the first coordinates. More...

Static Public Member Functions
static SpectralClustering	CreateLocalScaleFromNN (const SquareMatrixDouble &distance_matrix, size_t sigma_neighborhood=7, double epsilon=std::numeric_limits< double >::max())
	Clustering with local automatic scale. More...

static SpectralClustering	CreateGlobalScaleFromDimension (const SquareMatrixDouble &distance_matrix, size_t dimension, double epsilon=std::numeric_limits< double >::max())
	Clustering with global automatic scale. More...

static SpectralClustering	CreateGlobalScaleFromNN (const SquareMatrixDouble &distance_matrix, size_t sigma_neighborhood=1, double epsilon=std::numeric_limits< double >::max())
	Clustering with global automatic scale. More...

static SpectralClustering	CreateFixedScale (const SquareMatrixDouble &distance_matrix, double sigma, double epsilon=std::numeric_limits< double >::max())
	Clustering with global fixed scale. More...

Detailed Description

Spectral clustering.

Spectral clustering using Ng, Jordan & Weiss formula

Author: Yann LEYDIER

Date: September 2012

Version: 0.1

Definition at line 49 of file CRNSpectralClustering.h.

Constructor & Destructor Documentation

crn::SpectralClustering::SpectralClustering ( const SpectralClustering & )

delete

crn::SpectralClustering::SpectralClustering ( SpectralClustering && )

default

Member Function Documentation

SpectralClustering SpectralClustering::CreateFixedScale	(	const SquareMatrixDouble &	distance_matrix,
		double	sigma,
		double	epsilon = `std::numeric_limits<double>::max()`
	)

static

Clustering with global fixed scale.

Clustering with local scale

Exceptions

ExceptionInvalidArgument	sigma < 0
ExceptionDimension	empty distance matrix
ExceptionRuntime	too many iterations

Parameters

[in]	distance_matrix	the distance matrix of the data
[in]	sigma	the scale for Gaussian smoothing of the distance matrix
[in]	epsilon	the maximal distance between two elements (if distance > epsilon, then the elements cannot be in the same class)

Definition at line 146 of file CRNSpectralClustering.cpp.

SpectralClustering SpectralClustering::CreateGlobalScaleFromDimension	(	const SquareMatrixDouble &	distance_matrix,
		size_t	dimension,
		double	epsilon = `std::numeric_limits<double>::max()`
	)

static

Clustering with global automatic scale.

Clustering with global auto scale

Exceptions

ExceptionInvalidArgument	dimension <= 1
ExceptionDimension	empty distance matrix
ExceptionRuntime	too many iterations

Parameters

[in]	distance_matrix	the distance matrix of the data
[in]	dimension	the dimension of the original data
[in]	epsilon	the maximal distance between two elements (if distance > epsilon, then the elements cannot be in the same class)

Definition at line 119 of file CRNSpectralClustering.cpp.

SpectralClustering SpectralClustering::CreateGlobalScaleFromNN	(	const SquareMatrixDouble &	distance_matrix,
		size_t	sigma_neighborhood = `1`,
		double	epsilon = `std::numeric_limits<double>::max()`
	)

static

Clustering with global automatic scale.

Clustering with global auto scale

Exceptions

ExceptionInvalidArgument	neighborhood <= 1
ExceptionDimension	empty distance matrix
ExceptionRuntime	too many iterations

Parameters

[in]	distance_matrix	the distance matrix of the data
[in]	sigma_neighborhood	the number of neighbors used to compute the local sigma
[in]	epsilon	the maximal distance between two elements (if distance > epsilon, then the elements cannot be in the same class)

Definition at line 84 of file CRNSpectralClustering.cpp.

SpectralClustering SpectralClustering::CreateLocalScaleFromNN	(	const SquareMatrixDouble &	distance_matrix,
		size_t	sigma_neighborhood = `7`,
		double	epsilon = `std::numeric_limits<double>::max()`
	)

static

Clustering with local automatic scale.

Clustering with local auto scale

Exceptions

ExceptionInvalidArgument	neighborhood <= 1
ExceptionDimension	empty distance matrix
ExceptionRuntime	too many iterations

Parameters

[in]	distance_matrix	the distance matrix of the data
[in]	sigma_neighborhood	the number of neighbors used to compute the local sigma
[in]	epsilon	the maximal distance between two elements (if distance > epsilon, then the elements cannot be in the same class)

Definition at line 42 of file CRNSpectralClustering.cpp.

size_t SpectralClustering::EstimateClusterNumber ( double limit = 1.0 ) const

Estimates the number of clusters.

Estimates the number of clusters

Exceptions

ExceptionDomain limit is <0 or >1

Parameters

[in] limit the minimal eigenvalue to select. Normally it should be 1.0, but computational errors can make it necessary to use 0.9 for example.

Returns: the number of classes and optimal dimension of the projected data

Definition at line 206 of file CRNSpectralClustering.cpp.

const std::multimap<double, MatrixDouble>& crn::SpectralClustering::GetEigenpairs ( ) const

inlinenoexcept

Gets the data projected on each coordinates (higher eigenvalues are associated the to most significant coordinates)

Definition at line 67 of file CRNSpectralClustering.h.

std::vector< double > SpectralClustering::GetEigenvalues ( ) const

Returns the eigenvalues (sorted from highest to lowest)

Returns: the eigenvalues sorted from highest to lowest

Definition at line 192 of file CRNSpectralClustering.cpp.

SpectralClustering& crn::SpectralClustering::operator= ( const SpectralClustering & )

delete

SpectralClustering& crn::SpectralClustering::operator= ( SpectralClustering && )

default

std::vector< std::vector< double > > SpectralClustering::ProjectData	(	size_t	ncoordinates,
		bool	normalize
	)		const

Projects the data on the first coordinates.

Projects the data on the first coordinates

Exceptions

ExceptionDimension ncoordinates < 1

Parameters

[in]	ncoordinates	the dimension of the projection space
[in]	normalize	shall the elements be normalized on the unit circle?

Returns: a vector of projected elements

Definition at line 226 of file CRNSpectralClustering.cpp.

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

/home/travis/build/Liris-Pleiad/libcrn/src/libcrn/CRNAI/CRNSpectralClustering.h
/home/travis/build/Liris-Pleiad/libcrn/src/libcrn/CRNAI/CRNSpectralClustering.cpp

Public Member Functions

Static Public Member Functions

Detailed Description

Constructor & Destructor Documentation

Member Function Documentation