PrincipalComponentAnalysis Class Reference

#include <PrincipalComponentAnalysis.h>

Inheritance diagram for PrincipalComponentAnalysis:

Public Member Functions
	PrincipalComponentAnalysis ()
virtual	~PrincipalComponentAnalysis ()
bool	computeFeatureVector (const MatrixFloat &data, Float maxVariance=0.95, bool normData=false)
bool	computeFeatureVector (const MatrixFloat &data, UINT numPrincipalComponents, bool normData=false)
bool	project (const MatrixFloat &data, MatrixFloat &prjData)
bool	project (const VectorFloat &data, VectorFloat &prjData)
virtual bool	save (std::fstream &file) const
virtual bool	load (std::fstream &file)
bool	getTrained () const
bool	getNormData () const
UINT	getNumInputDimensions () const
UINT	getNumPrincipalComponents () const
Float	getMaxVariance () const
VectorFloat	getMeanVector () const
VectorFloat	getStdDevVector () const
VectorFloat	getComponentWeights () const
VectorFloat	getEigenValues () const
virtual bool	print (std::string title="") const
MatrixFloat	getEigenVectors () const
bool	setModel (const VectorFloat &mean, const MatrixFloat &eigenvectors)
Public Member Functions inherited from MLBase
	MLBase (void)
virtual	~MLBase (void)
bool	copyMLBaseVariables (const MLBase *mlBase)
virtual bool	train (ClassificationData trainingData)
virtual bool	train_ (ClassificationData &trainingData)
virtual bool	train (RegressionData trainingData)
virtual bool	train_ (RegressionData &trainingData)
virtual bool	train (TimeSeriesClassificationData trainingData)
virtual bool	train_ (TimeSeriesClassificationData &trainingData)
virtual bool	train (ClassificationDataStream trainingData)
virtual bool	train_ (ClassificationDataStream &trainingData)
virtual bool	train (UnlabelledData trainingData)
virtual bool	train_ (UnlabelledData &trainingData)
virtual bool	train (MatrixFloat data)
virtual bool	train_ (MatrixFloat &data)
virtual bool	predict (VectorFloat inputVector)
virtual bool	predict_ (VectorFloat &inputVector)
virtual bool	predict (MatrixFloat inputMatrix)
virtual bool	predict_ (MatrixFloat &inputMatrix)
virtual bool	map (VectorFloat inputVector)
virtual bool	map_ (VectorFloat &inputVector)
virtual bool	reset ()
virtual bool	clear ()
virtual bool	print () const
virtual bool	save (const std::string filename) const
virtual bool	load (const std::string filename)
	GRT_DEPRECATED_MSG ("saveModelToFile(std::string filename) is deprecated, use save(std::string filename) instead", virtual bool saveModelToFile(std::string filename) const )
	GRT_DEPRECATED_MSG ("saveModelToFile(std::fstream &file) is deprecated, use save(std::fstream &file) instead", virtual bool saveModelToFile(std::fstream &file) const )
	GRT_DEPRECATED_MSG ("loadModelFromFile(std::string filename) is deprecated, use load(std::string filename) instead", virtual bool loadModelFromFile(std::string filename))
	GRT_DEPRECATED_MSG ("loadModelFromFile(std::fstream &file) is deprecated, use load(std::fstream &file) instead", virtual bool loadModelFromFile(std::fstream &file))
virtual bool	getModel (std::ostream &stream) const
Float	scale (const Float &x, const Float &minSource, const Float &maxSource, const Float &minTarget, const Float &maxTarget, const bool constrain=false)
virtual std::string	getModelAsString () const
DataType	getInputType () const
DataType	getOutputType () const
UINT	getBaseType () const
UINT	getNumInputFeatures () const
UINT	getNumInputDimensions () const
UINT	getNumOutputDimensions () const
UINT	getMinNumEpochs () const
UINT	getMaxNumEpochs () const
UINT	getValidationSetSize () const
UINT	getNumTrainingIterationsToConverge () const
Float	getMinChange () const
Float	getLearningRate () const
Float	getRootMeanSquaredTrainingError () const
Float	getTotalSquaredTrainingError () const
Float	getValidationSetAccuracy () const
VectorFloat	getValidationSetPrecision () const
VectorFloat	getValidationSetRecall () const
bool	getUseValidationSet () const
bool	getRandomiseTrainingOrder () const
bool	getTrained () const
bool	getModelTrained () const
bool	getScalingEnabled () const
bool	getIsBaseTypeClassifier () const
bool	getIsBaseTypeRegressifier () const
bool	getIsBaseTypeClusterer () const
bool	enableScaling (const bool useScaling)
bool	setMaxNumEpochs (const UINT maxNumEpochs)
bool	setMinNumEpochs (const UINT minNumEpochs)
bool	setMinChange (const Float minChange)
bool	setLearningRate (const Float learningRate)
bool	setUseValidationSet (const bool useValidationSet)
bool	setValidationSetSize (const UINT validationSetSize)
bool	setRandomiseTrainingOrder (const bool randomiseTrainingOrder)
bool	setTrainingLoggingEnabled (const bool loggingEnabled)
bool	registerTrainingResultsObserver (Observer< TrainingResult > &observer)
bool	registerTestResultsObserver (Observer< TestInstanceResult > &observer)
bool	removeTrainingResultsObserver (const Observer< TrainingResult > &observer)
bool	removeTestResultsObserver (const Observer< TestInstanceResult > &observer)
bool	removeAllTrainingObservers ()
bool	removeAllTestObservers ()
bool	notifyTrainingResultsObservers (const TrainingResult &data)
bool	notifyTestResultsObservers (const TestInstanceResult &data)
MLBase *	getMLBasePointer ()
const MLBase *	getMLBasePointer () const
Vector< TrainingResult >	getTrainingResults () const
Public Member Functions inherited from GRTBase
	GRTBase (void)
virtual	~GRTBase (void)
bool	copyGRTBaseVariables (const GRTBase *GRTBase)
std::string	getClassType () const
std::string	getLastWarningMessage () const
std::string	getLastErrorMessage () const
std::string	getLastInfoMessage () const
bool	setInfoLoggingEnabled (const bool loggingEnabled)
bool	setWarningLoggingEnabled (const bool loggingEnabled)
bool	setErrorLoggingEnabled (const bool loggingEnabled)
GRTBase *	getGRTBasePointer ()
const GRTBase *	getGRTBasePointer () const
Public Member Functions inherited from Observer< TrainingResult >
virtual void	notify (const TrainingResult &data)
Public Member Functions inherited from Observer< TestInstanceResult >
virtual void	notify (const TestInstanceResult &data)

Protected Types
enum	AnalysisMode { MAX_VARIANCE =0, MAX_NUM_PCS }

Protected Member Functions
bool	computeFeatureVector_ (const MatrixFloat &data, UINT analysisMode)
Protected Member Functions inherited from MLBase
bool	saveBaseSettingsToFile (std::fstream &file) const
bool	loadBaseSettingsFromFile (std::fstream &file)
Protected Member Functions inherited from GRTBase
Float	SQR (const Float &x) const

Protected Attributes
bool	normData
UINT	numPrincipalComponents
Float	maxVariance
VectorFloat	mean
VectorFloat	stdDev
VectorFloat	componentWeights
VectorFloat	eigenvalues
Vector< IndexedDouble >	sortedEigenvalues
MatrixFloat	eigenvectors
Protected Attributes inherited from MLBase
bool	trained
bool	useScaling
DataType	inputType
DataType	outputType
UINT	baseType
UINT	numInputDimensions
UINT	numOutputDimensions
UINT	numTrainingIterationsToConverge
UINT	minNumEpochs
UINT	maxNumEpochs
UINT	validationSetSize
Float	learningRate
Float	minChange
Float	rootMeanSquaredTrainingError
Float	totalSquaredTrainingError
Float	validationSetAccuracy
bool	useValidationSet
bool	randomiseTrainingOrder
VectorFloat	validationSetPrecision
VectorFloat	validationSetRecall
Random	random
std::vector< TrainingResult >	trainingResults
TrainingResultsObserverManager	trainingResultsObserverManager
TestResultsObserverManager	testResultsObserverManager
Protected Attributes inherited from GRTBase
std::string	classType
DebugLog	debugLog
ErrorLog	errorLog
InfoLog	infoLog
TrainingLog	trainingLog
TestingLog	testingLog
WarningLog	warningLog

Additional Inherited Members
Public Types inherited from MLBase
enum	BaseTypes { BASE_TYPE_NOT_SET =0, CLASSIFIER, REGRESSIFIER, CLUSTERER }
Static Public Member Functions inherited from GRTBase
static std::string	getGRTVersion (bool returnRevision=true)
static std::string	getGRTRevison ()

Detailed Description

GRT MIT License Copyright (c) <2012> <Nicholas Gillian, Media Lab, MIT>

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

Definition at line 51 of file PrincipalComponentAnalysis.h.

Constructor & Destructor Documentation

GRT_BEGIN_NAMESPACE PrincipalComponentAnalysis::PrincipalComponentAnalysis

(

)

Default constructor.

Definition at line 31 of file PrincipalComponentAnalysis.cpp.

PrincipalComponentAnalysis::~PrincipalComponentAnalysis ( )

virtual

Default destructor.

Definition at line 43 of file PrincipalComponentAnalysis.cpp.

Member Function Documentation

bool PrincipalComponentAnalysis::computeFeatureVector	(	const MatrixFloat &	data,
		Float	maxVariance = 0.95,
		bool	normData = false
	)

Runs the principal component analysis algorithm on the input data and builds the resulting feature vector so new data can be projected onto the principal subspace (using the project function). The number of principal components is automatically computed by selecting the minimum number of components that reach the maxVariance value. This should be a value between [0 1], the default value of 0.95 represents 95% of the variance in the original dataset.

Parameters

data	a matrix containing the data from which the principal components will be computed. This should be an [M N] matrix, where M==samples and N==dimensions.
maxVariance	sets the variance that should represented by the top K principal components. This should be a value between [0 1]. Default value=0.95
normData	sets if the data will be z-normalized before running the PCA algorithm. Default value=false

Returns

returns true if the principal components of the input matrix could be computed, false otherwise value

bool PrincipalComponentAnalysis::computeFeatureVector	(	const MatrixFloat &	data,
		UINT	numPrincipalComponents,
		bool	normData = false
	)

Runs the principal component analysis algorithm on the input data and builds the resulting feature vector so new data can be projected onto the principal subspace (using the project function). The number of principal components should be set be the user and must be less than or equal to the number of dimensions in the input data.

Parameters

data	a matrix containing the data from which the principal components will be computed. This should be an [M N] matrix, where M==samples and N==dimensions
numPrincipalComponents	sets the number of principal components. This must be a value be less than or equal to the number of dimensions in the input data
normData	sets if the data will be z-normalized before running the PCA algorithm. Default value=false

Returns

returns true if the principal components of the input matrix could be computed, false otherwise value

Definition at line 54 of file PrincipalComponentAnalysis.cpp.

VectorFloat PrincipalComponentAnalysis::getComponentWeights ( ) const

inline

Returns the weights for each principal component, these weights sum to 1.

Returns

returns a vector of the weights for each principal component, these weights sum to 1

Definition at line 178 of file PrincipalComponentAnalysis.h.

VectorFloat PrincipalComponentAnalysis::getEigenValues ( ) const

inline

Returns the raw eigen values (these are not sorted).

Returns

returns a vector of the raw eigen values

Definition at line 184 of file PrincipalComponentAnalysis.h.

MatrixFloat PrincipalComponentAnalysis::getEigenVectors

(

)

const

Returns a matrix containing the eigen vectors.

Returns

returns a matrix containing the raw eigen vectors

Definition at line 455 of file PrincipalComponentAnalysis.cpp.

Float PrincipalComponentAnalysis::getMaxVariance ( ) const

inline

Returns the maxVariance parameter, set by the user when the computeFeatureVector was called. returns the maxVariance parameter, set by the user when the computeFeatureVector was called

Definition at line 157 of file PrincipalComponentAnalysis.h.

VectorFloat PrincipalComponentAnalysis::getMeanVector ( ) const

inline

Returns the mean shift vector, computed during the computeFeatureVector function. New data will be subtracted by this value before it is projected onto the principal subspace.

Returns

returns the mean shift vector, computed during the computeFeatureVector function

Definition at line 164 of file PrincipalComponentAnalysis.h.

bool PrincipalComponentAnalysis::getNormData ( ) const

inline

Returns true if z-normalization is being applied to new data.

Returns

returns true if the normData is true, false otherwise

Definition at line 139 of file PrincipalComponentAnalysis.h.

UINT PrincipalComponentAnalysis::getNumInputDimensions ( ) const

inline

Returns the number of input dimensions in the original input data.

Returns

returns the numInputDimensions parameter.

Definition at line 145 of file PrincipalComponentAnalysis.h.

UINT PrincipalComponentAnalysis::getNumPrincipalComponents ( ) const

inline

Returns the number of principal components that was required to reach the maxVariance parameter.

Returns

returns the number of principal components that was required to reach the maxVariance parameter

Definition at line 151 of file PrincipalComponentAnalysis.h.

VectorFloat PrincipalComponentAnalysis::getStdDevVector ( ) const

inline

Returns the standard deviation vector that is used to normalize new data, this is computed during the computeFeatureVector function. This is only used id the normData parameter is true. If true, new data will be z-normalized by this value before it is projected onto the principal subspace.

Returns

returns the stdDev vector, computed during the computeFeatureVector function

Definition at line 172 of file PrincipalComponentAnalysis.h.

bool PrincipalComponentAnalysis::getTrained ( ) const

inline

Returns true if the module was trained.

Returns

returns true if the module has been trained, false otherwise

Definition at line 133 of file PrincipalComponentAnalysis.h.

bool PrincipalComponentAnalysis::load ( std::fstream &  file )

virtual

This loads a trained PCA model from a file.

Parameters

file	a reference to the file the model will be loaded from

Returns

returns true if the model was loaded successfully, false otherwise

Reimplemented from MLBase.

Definition at line 313 of file PrincipalComponentAnalysis.cpp.

bool PrincipalComponentAnalysis::print ( std::string  title = "" ) const

virtual

A helper function that prints the PCA info. If the user sets the title string, then this will be written in addition with the PCA data.

Definition at line 430 of file PrincipalComponentAnalysis.cpp.

bool PrincipalComponentAnalysis::project	(	const MatrixFloat &	data,
		MatrixFloat &	prjData
	)

Projects the input data matrix onto the principal subspace. The new projected data will be stored in the prjData matrix. The computeFeatureVector function should have been called at least once before this function is called. The number of the columns in the data matrix must match the numInputDimensions parameter. The function will return true if the projection was successful, false otherwise.

Parameters

data	The data that should be projected onto the principal subspace. This should be an [M N] matrix, where N must equal the numInputDimensions value (there are no restrictions on M).
prjData	A matrix into which the projected data will be stored. This matrix will be resized to [M K], where M is the number of rows in the data matrix and K is the numPrincipalComponents.

Returns

returns true if the projection was successful, false otherwise

Definition at line 179 of file PrincipalComponentAnalysis.cpp.

bool PrincipalComponentAnalysis::project	(	const VectorFloat &	data,
		VectorFloat &	prjData
	)

Projects the input data vector onto the principal subspace. The new projected data will be stored in the prjData vector. The computeFeatureVector function should have been called at least once before this function is called. The size of the data vector must match the numInputDimensions parameter. The function will return true if the projection was successful, false otherwise.

Parameters

data	The data that should be projected onto the principal subspace. This should be an N-dimensional vector, where N must equal the numInputDimensions value.
prjData	A vector into which the projected data will be stored. This vector will be resized to K, where K is the numPrincipalComponents.

Returns

returns true if the projection was successful, false otherwise

Definition at line 218 of file PrincipalComponentAnalysis.cpp.

bool PrincipalComponentAnalysis::save ( std::fstream &  file ) const

virtual

This saves the trained PCA model to a file.

Parameters

file	a reference to the file the model will be saved to

Returns

returns true if the model was saved successfully, false otherwise

Reimplemented from MLBase.

Definition at line 255 of file PrincipalComponentAnalysis.cpp.

---

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

• PrincipalComponentAnalysis.h
• PrincipalComponentAnalysis.cpp