Class to perform Principal Componant Analysis. More...

#include <CRNPCA.h>

Inheritance diagram for crn::PCA:

Collaboration diagram for crn::PCA:

Public Member Functions
	PCA (const MatrixDouble &data, bool data_reduction_flag=true)
	Constructor. More...

	PCA (const std::vector< std::vector< double > > &data, bool data_reduction_flag=true)
	Constructor. More...

	PCA (const std::vector< std::vector< double > > &data, const std::vector< size_t > &cards, bool data_reduction_flag=true)
	Constructor from weighted vectors. More...

	PCA (const std::map< std::vector< double >, size_t > &data, bool data_reduction_flag=true)
	Constructor from map. More...

template<typename ITER >
	PCA (ITER begin, ITER end, bool data_reduction_flag=true)
	Constructor from iterators. More...

	PCA (const PCA &)=default
	Constructor. More...

	PCA (PCA &&)=default

virtual	~PCA () override=default
	Destructor. More...

PCA &	operator= (const PCA &)=default

PCA &	operator= (PCA &&)=default

size_t	GetDimension () const noexcept
	Returns the dimension of feature space. More...

const std::vector< double > &	GetMeans () const noexcept
	Returns the mean values computed for sample data. More...

double	GetMean (size_t d) const
	Returns the mean value of d-th feature computed for sample data. More...

const std::vector< double > &	GetDeviations () const noexcept
	Returns the deviation computed for sample data. More...

double	GetDeviation (size_t d) const
	Returns the deviation of d-th feature computed for sample data. More...

const std::multimap< double, MatrixDouble > &	GetEigensystem () const noexcept
	Returns the eigensystem computed on covariance matrix of last sample data. More...

MatrixDouble	Transform (const MatrixDouble &patterns, size_t nb_features=1u) const
	Apply transform to given patterns. More...

std::vector< std::vector < double > >	Transform (const std::vector< std::vector< double > > &data, const size_t nb_features=0u) const
	Apply transform to given patterns. More...

std::vector< std::vector < double > >	ReverseTransform (const std::vector< std::vector< double > > &data) const
	Apply reverse transform to get given patterns' pre-images. More...

void	Deserialize (xml::Element &el)

xml::Element	Serialize (xml::Element &parent) const

	PCA (xml::Element &el)

Public Member Functions inherited from crn::Object
virtual	~Object ()=default

Detailed Description

Class to perform Principal Componant Analysis.

Principal Componant Analysis

Author: Jean DUONG

Date: September 2009

Version: 0.1

Definition at line 39 of file CRNPCA.h.

Constructor & Destructor Documentation

PCA::PCA	(	const MatrixDouble &	data,
		bool	data_reduction_flag = `true`
	)

Constructor.

Constructor

Parameters

[in]	data	matrix of double numbers in which patterns are stored as rows
[in]	data_reduction_flag	flag to indicate if the data have to be reduced

Definition at line 41 of file CRNPCA.cpp.

PCA::PCA	(	const std::vector< std::vector< double > > &	data,
		bool	data_reduction_flag = `true`
	)

Constructor.

Constructor

Parameters

[in]	data	vector of vectors of double numbers in which patterns are stored
[in]	data_reduction_flag	flag to indicate if the data have to be reduced

Definition at line 140 of file CRNPCA.cpp.

PCA::PCA	(	const std::vector< std::vector< double > > &	data,
		const std::vector< size_t > &	cards,
		bool	data_reduction_flag = `true`
	)

Constructor from weighted vectors.

Constructor

Parameters

[in]	data	collection of vectors of double numbers
[in]	cards	pattern cardinals
[in]	data_reduction_flag	flag to indicate if the data have to be reduced

Definition at line 247 of file CRNPCA.cpp.

PCA::PCA	(	const std::map< std::vector< double >, size_t > &	data,
		bool	data_reduction_flag = `true`
	)

Constructor from map.

Constructor

Parameters

[in]	data	map of vectors of double numbers with cardinals
[in]	data_reduction_flag	flag to indicate if the data have to be reduced

Definition at line 359 of file CRNPCA.cpp.

template<typename ITER >

crn::PCA::PCA	(	ITER	begin,
		ITER	end,
		bool	data_reduction_flag = `true`
	)

Constructor from iterators.

Constructor

Performs principal component analysis over patterns given as a value-cardinal pairs

Parameters

[in]	begin	iterator pointing at the begining of the data collection
[in]	end	ending iterator
[in]	data_reduction_flag	flag to indicate if data should be reduced for PCA estimate

Definition at line 114 of file CRNPCA.h.

crn::PCA::PCA ( const PCA & )

default

Constructor.

crn::PCA::PCA ( PCA && )

default

virtual crn::PCA::~PCA ( )

overridevirtualdefault

Destructor.

crn::PCA::PCA ( xml::Element & el )

inline

Definition at line 97 of file CRNPCA.h.

Member Function Documentation

void PCA::Deserialize ( xml::Element & el )

Unsafe load from XML node

Author: Jean DUONG

Exceptions

ExceptionInvalidArgument	not a UnivariateGaussianMixture
ExceptionNotFound	cannot find attribute
ExceptionDomain	wrong attribute

Parameters

[in] el the element to load

Definition at line 713 of file CRNPCA.cpp.

double PCA::GetDeviation ( size_t d ) const

Returns the deviation of d-th feature computed for sample data.

Returns the deviation of d-th feature

Exceptions

ExceptionDimension incompatible patter dimensions

Parameters

[in] d index of desired value

Returns: double value

Definition at line 539 of file CRNPCA.cpp.

const std::vector<double>& crn::PCA::GetDeviations ( ) const

inlinenoexcept

Returns the deviation computed for sample data.

Definition at line 69 of file CRNPCA.h.

size_t crn::PCA::GetDimension ( ) const

inlinenoexcept

Returns the dimension of feature space.

Definition at line 62 of file CRNPCA.h.

const std::multimap<double, MatrixDouble>& crn::PCA::GetEigensystem ( ) const

inlinenoexcept

Returns the eigensystem computed on covariance matrix of last sample data.

Definition at line 74 of file CRNPCA.h.

double PCA::GetMean ( size_t d ) const

Returns the mean value of d-th feature computed for sample data.

Returns the mean value of d-th feature

Exceptions

ExceptionDimension incompatible patter dimensions

Parameters

[in] d index of desired value

Returns: double value

Definition at line 524 of file CRNPCA.cpp.

const std::vector<double>& crn::PCA::GetMeans ( ) const

inlinenoexcept

Returns the mean values computed for sample data.

Definition at line 64 of file CRNPCA.h.

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

default

PCA& crn::PCA::operator= ( PCA && )

default

std::vector< std::vector< double > > PCA::ReverseTransform ( const std::vector< std::vector< double > > & data ) const

Apply reverse transform to get given patterns' pre-images.

Apply reverse transform to get given patterns' pre-images. No dimension change allowed for this version.

Parameters

[in] data vector of multivariate patterns

Returns: a vector containing projected multivariate patterns

Definition at line 663 of file CRNPCA.cpp.

xml::Element PCA::Serialize ( xml::Element & parent ) const

Unsafe save

Author: Jean DUONG

Parameters

[in] parent the parent element to which we will add the new element

Returns: The newly created element

Definition at line 773 of file CRNPCA.cpp.

MatrixDouble PCA::Transform	(	const MatrixDouble &	patterns,
		size_t	nb_features = `1u`
	)		const

Apply transform to given patterns.

Perform projection of given patterns using current PCA attributes. Patterns are translated to make centroid become the center of space, then projected on pricipal componants from most to least significant ones.

Exceptions

ExceptionDimension	incompatible patter dimensions
ExceptionDomain	incompatible output dimension

Parameters

[in]	patterns	matrix of double numbers in which patterns are stored
[in]	nb_features	the number of principal componants used for projection

Returns: a matrix containing projected patterns

Definition at line 560 of file CRNPCA.cpp.

std::vector< std::vector< double > > PCA::Transform	(	const std::vector< std::vector< double > > &	data,
		const size_t	nb_features = `0u`
	)		const

Apply transform to given patterns.

Perform projection of given patterns using current PCA attributes. Patterns are translated to make centroid become the center of space, then projected on principal componants from most to least significant ones.

Parameters

[in]	data	vector of multivariate patterns
[in]	nb_features	the number of principal componants used for projection

Returns: a vector containing projected multivariate patterns

Definition at line 614 of file CRNPCA.cpp.

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

/home/travis/build/Liris-Pleiad/libcrn/src/libcrn/CRNStatistics/CRNPCA.h
/home/travis/build/Liris-Pleiad/libcrn/src/libcrn/CRNStatistics/CRNPCA.cpp

Public Member Functions

Detailed Description

Constructor & Destructor Documentation

Member Function Documentation