Inheritance diagram for MLP:

Public Types
enum	TrainingAlgorithm { ONLINE_GRADIENT_DESCENT =0 }

Public Types inherited from Regressifier
typedef std::map< std::string, Regressifier ()() >	StringRegressifierMap

Public Types inherited from MLBase
enum	BaseType { BASE_TYPE_NOT_SET =0, CLASSIFIER, REGRESSIFIER, CLUSTERER, PRE_PROCSSING, POST_PROCESSING, FEATURE_EXTRACTION, CONTEXT }

Public Member Functions
	MLP ()

	MLP (const MLP &rhs)

virtual	~MLP ()

MLP &	operator= (const MLP &rhs)

virtual bool	deepCopyFrom (const Regressifier *regressifier)

virtual bool	train_ (ClassificationData &trainingData)

virtual bool	train_ (RegressionData &trainingData)

virtual bool	predict_ (VectorFloat &inputVector)

virtual bool	clear ()

virtual bool	print () const

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

virtual bool	load (std::fstream &file)

UINT	getNumClasses () const

bool	init (const UINT numInputNeurons, const UINT numHiddenNeurons, const UINT numOutputNeurons)

bool	init (const UINT numInputNeurons, const UINT numHiddenNeurons, const UINT numOutputNeurons, const Neuron::Type inputLayerActivationFunction, const Neuron::Type hiddenLayerActivationFunction, const Neuron::Type outputLayerActivationFunction)

void	printNetwork () const

bool	checkForNAN () const

std::string	activationFunctionToString (const Neuron::Type activationFunction) const

Neuron::Type	activationFunctionFromString (const std::string activationName) const

bool	validateActivationFunction (const Neuron::Type avactivationFunction) const

UINT	getNumInputNeurons () const

UINT	getNumHiddenNeurons () const

UINT	getNumOutputNeurons () const

UINT	getNumRandomTrainingIterations () const

Neuron::Type	getInputLayerActivationFunction () const

Neuron::Type	getHiddenLayerActivationFunction () const

Neuron::Type	getOutputLayerActivationFunction () const

Float	getTrainingRate () const

Float	getMomentum () const

Float	getGamma () const

Float	getTrainingError () const

bool	getClassificationModeActive () const

bool	getRegressionModeActive () const

Vector< Neuron >	getInputLayer () const

Vector< Neuron >	getHiddenLayer () const

Vector< Neuron >	getOutputLayer () const

Vector< VectorFloat >	getTrainingLog () const

bool	getNullRejectionEnabled () const

Float	getNullRejectionCoeff () const

Float	getNullRejectionThreshold () const

Float	getMaximumLikelihood () const

VectorFloat	getClassLikelihoods () const

VectorFloat	getClassDistances () const

UINT	getPredictedClassLabel () const

bool	setInputLayerActivationFunction (const Neuron::Type activationFunction)

bool	setHiddenLayerActivationFunction (const Neuron::Type activationFunction)

bool	setOutputLayerActivationFunction (const Neuron::Type activationFunction)

bool	setTrainingRate (const Float trainingRate)

bool	setMomentum (const Float momentum)

bool	setGamma (const Float gamma)

	GRT_DEPRECATED_MSG ("setNumRandomTrainingIterations() is deprecated, use setNumRestarts() instead", bool setNumRandomTrainingIterations(const UINT numRandomTrainingIterations))

bool	setNullRejection (const bool useNullRejection)

bool	setNullRejectionCoeff (const Float nullRejectionCoeff)

Public Member Functions inherited from Regressifier
	Regressifier (const std::string &id="")

virtual	~Regressifier (void)

bool	copyBaseVariables (const Regressifier *regressifier)

virtual bool	reset () override

VectorFloat	getRegressionData () const

Vector< MinMax >	getInputRanges () const

Vector< MinMax >	getOutputRanges () const

Regressifier *	deepCopy () const

const Regressifier &	getBaseRegressifier () const

Regressifier *	create () const

	GRT_DEPRECATED_MSG ("createNewInstance is deprecated, use create() instead.", Regressifier *createNewInstance() const )

	GRT_DEPRECATED_MSG ("createInstanceFromString(id) is deprecated, use create(id) instead.", static Regressifier *createInstanceFromString(const std::string &id))

	GRT_DEPRECATED_MSG ("getRegressifierType is deprecated, use getId() instead", std::string getRegressifierType() const )

Public Member Functions inherited from MLBase
	MLBase (const std::string &id="", const BaseType type=BASE_TYPE_NOT_SET)

virtual	~MLBase (void)

bool	copyMLBaseVariables (const MLBase *mlBase)

virtual bool	train (ClassificationData trainingData)

virtual bool	train (RegressionData trainingData)

virtual bool	train (RegressionData trainingData, RegressionData validationData)

virtual bool	train_ (RegressionData &trainingData, RegressionData &validationData)

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 (MatrixFloat inputMatrix)

virtual bool	predict_ (MatrixFloat &inputMatrix)

virtual bool	map (VectorFloat inputVector)

virtual bool	map_ (VectorFloat &inputVector)

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(const std::string &filename) instead", virtual bool saveModelToFile(const 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(const std::string &filename) instead", virtual bool loadModelFromFile(const 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

virtual std::string	getModelAsString () const

DataType	getInputType () const

DataType	getOutputType () const

BaseType	getType () const

UINT	getNumInputFeatures () const

UINT	getNumInputDimensions () const

UINT	getNumOutputDimensions () const

UINT	getMinNumEpochs () const

UINT	getMaxNumEpochs () const

UINT	getBatchSize () const

UINT	getNumRestarts () const

UINT	getValidationSetSize () const

UINT	getNumTrainingIterationsToConverge () const

Float	getMinChange () const

Float	getLearningRate () const

Float	getRMSTrainingError () const

	GRT_DEPRECATED_MSG ("getRootMeanSquaredTrainingError() is deprecated, use getRMSTrainingError() instead", Float getRootMeanSquaredTrainingError() const )

Float	getTotalSquaredTrainingError () const

Float	getRMSValidationError () const

Float	getValidationSetAccuracy () const

VectorFloat	getValidationSetPrecision () const

VectorFloat	getValidationSetRecall () const

bool	getUseValidationSet () const

bool	getRandomiseTrainingOrder () const

bool	getTrained () const

	GRT_DEPRECATED_MSG ("getModelTrained() is deprecated, use getTrained() instead", bool getModelTrained() const )

bool	getConverged () const

bool	getScalingEnabled () const

bool	getIsBaseTypeClassifier () const

bool	getIsBaseTypeRegressifier () const

bool	getIsBaseTypeClusterer () const

bool	getTrainingLoggingEnabled () const

bool	getTestingLoggingEnabled () const

bool	enableScaling (const bool useScaling)

bool	setMaxNumEpochs (const UINT maxNumEpochs)

bool	setBatchSize (const UINT batchSize)

bool	setMinNumEpochs (const UINT minNumEpochs)

bool	setNumRestarts (const UINT numRestarts)

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	setTestingLoggingEnabled (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 (const std::string &id="")

virtual	~GRTBase (void)

bool	copyGRTBaseVariables (const GRTBase *GRTBase)

	GRT_DEPRECATED_MSG ("getClassType is deprecated, use getId() instead!", std::string getClassType() const )

std::string	getId () 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)

bool	setDebugLoggingEnabled (const bool loggingEnabled)

GRTBase *	getGRTBasePointer ()

const GRTBase *	getGRTBasePointer () const

Float	scale (const Float &x, const Float &minSource, const Float &maxSource, const Float &minTarget, const Float &maxTarget, const bool constrain=false)

Float	SQR (const Float &x) 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)

Static Public Member Functions
static std::string	getId ()

Static Public Member Functions inherited from Regressifier
static Vector< std::string >	getRegisteredRegressifiers ()

static Regressifier *	create (const std::string &id)

Static Public Member Functions inherited from GRTBase
static std::string	getGRTVersion (bool returnRevision=true)

static std::string	getGRTRevison ()

Protected Member Functions
bool	isNAN (const Float &v) const

bool	setOutputTargets ()

bool	trainModel (RegressionData &trainingData)

bool	trainOnlineGradientDescentClassification (const RegressionData &trainingData, const RegressionData &validationData)

bool	trainOnlineGradientDescentRegression (const RegressionData &trainingData, const RegressionData &validationData)

bool	loadLegacyModelFromFile (std::fstream &file)

Float	back_prop (const VectorFloat &inputVector, const VectorFloat &targetVector, const Float alpha, const Float beta)

VectorFloat	feedforward (VectorFloat data)

void	feedforward (const VectorFloat &data, VectorFloat &inputNeuronsOutput, VectorFloat &hiddenNeuronsOutput, VectorFloat &outputNeuronsOutput)

Protected Member Functions inherited from Regressifier
bool	saveBaseSettingsToFile (std::fstream &file) const

bool	loadBaseSettingsFromFile (std::fstream &file)

Protected Member Functions inherited from MLBase
bool	saveBaseSettingsToFile (std::fstream &file) const

bool	loadBaseSettingsFromFile (std::fstream &file)

Protected Attributes
UINT	numInputNeurons

UINT	numHiddenNeurons

UINT	numOutputNeurons

Neuron::Type	inputLayerActivationFunction

Neuron::Type	hiddenLayerActivationFunction

Neuron::Type	outputLayerActivationFunction

UINT	trainingMode

Float	momentum

Float	gamma

Float	trainingError

MinMax	outputTargets

bool	initialized

Vector< Neuron >	inputLayer

Vector< Neuron >	hiddenLayer

Vector< Neuron >	outputLayer

Vector< VectorFloat >	trainingErrorLog

bool	classificationModeActive

bool	useNullRejection

UINT	predictedClassLabel

Float	nullRejectionThreshold

Float	nullRejectionCoeff

Float	maxLikelihood

VectorFloat	classLikelihoods

VectorFloat	inputNeuronsOutput

VectorFloat	hiddenNeuronsOutput

VectorFloat	outputNeuronsOutput

VectorFloat	deltaO

VectorFloat	deltaH

Protected Attributes inherited from Regressifier
std::string	regressifierType

VectorFloat	regressionData

Vector< MinMax >	inputVectorRanges

Vector< MinMax >	targetVectorRanges

Protected Attributes inherited from MLBase
bool	trained

bool	useScaling

bool	converged

DataType	inputType

DataType	outputType

BaseType	baseType

UINT	numInputDimensions

UINT	numOutputDimensions

UINT	numTrainingIterationsToConverge

UINT	minNumEpochs

UINT	maxNumEpochs

UINT	batchSize

UINT	validationSetSize

UINT	numRestarts

Float	learningRate

Float	minChange

Float	rmsTrainingError

Float	rmsValidationError

Float	totalSquaredTrainingError

Float	validationSetAccuracy

bool	useValidationSet

bool	randomiseTrainingOrder

VectorFloat	validationSetPrecision

VectorFloat	validationSetRecall

Random	random

Vector< TrainingResult >	trainingResults

TrainingResultsObserverManager	trainingResultsObserverManager

TestResultsObserverManager	testResultsObserverManager

TrainingLog	trainingLog

TestingLog	testingLog

Protected Attributes inherited from GRTBase
std::string	classId
	Stores the name of the class (e.g., MinDist)

DebugLog	debugLog

ErrorLog	errorLog

InfoLog	infoLog

WarningLog	warningLog

Additional Inherited Members
Static Protected Member Functions inherited from Regressifier
static StringRegressifierMap *	getMap ()

Detailed Description

Examples:: RegressionModulesExamples/MLPRegressionExample/MLPRegressionExample.cpp.

Definition at line 42 of file MLP.h.

Constructor & Destructor Documentation

MLP::MLP ( )

Default Constructor

Definition at line 34 of file MLP.cpp.

MLP::MLP ( const MLP & rhs )

Copy Constructor

Parameters

rhs	another instance of a MLP that will be copied to this instance

Definition at line 58 of file MLP.cpp.

MLP::~MLP ( )

virtual

Default Destructor

Definition at line 63 of file MLP.cpp.

Member Function Documentation

Neuron::Type MLP::activationFunctionFromString ( const std::string activationName ) const

Gets the activation function value from a string.

Parameters

activationName the activation function as a string

Returns: returns the activation function

Definition at line 1623 of file MLP.cpp.

std::string MLP::activationFunctionToString ( const Neuron::Type activationFunction ) const

Gets a string representation of the activation function

Parameters

activationFunction the activation function you want to convert to a string

Returns: returns a string representation of the activation function, returns UNKNOWN if the activation function is invalid

Definition at line 1599 of file MLP.cpp.

Float MLP::back_prop	(	const VectorFloat &	inputVector,
		const VectorFloat &	targetVector,
		const Float	alpha,
		const Float	beta
	)

protected

Performs one round of back propagation, using the training example and target Vector

Parameters

inputVector	the input Vector to use for back propagation
targetVector	the target Vector to use for back propagation
alpha	the training rate
beta	the momentum

Returns: returns the squared error for the current training example

Definition at line 851 of file MLP.cpp.

bool MLP::checkForNAN ( ) const

Checks if there are any NAN values in any of the layers.

Returns: returns true if there are any NAN values in any of the layers, false otherwise

Definition at line 1039 of file MLP.cpp.

bool MLP::clear ( )

virtual

Clears any previous model or settings.

Returns: returns true if the MLP was cleared, false otherwise

Reimplemented from Regressifier.

Definition at line 291 of file MLP.cpp.

bool MLP::deepCopyFrom ( const Regressifier * regressifier )

virtual

This is required for the Gesture Recognition Pipeline for when the pipeline.setRegressifier(...) method is called. It clones the data from the Base Class Regressifier pointer (which should be pointing to an MLP instance) into this instance

Parameters

regressifier a pointer to the Regressifier Base Class, this should be pointing to another MLP instance

Returns: returns true if the clone was successfull, false otherwise

Reimplemented from Regressifier.

Definition at line 103 of file MLP.cpp.

VectorFloat MLP::feedforward ( VectorFloat data )

protected

Performs the feedforward step using the current model and the input training example.

Parameters

data	the input Vector to use for the feedforward

Returns: returns a new Vector of the results from the feedforward step

Definition at line 924 of file MLP.cpp.

void MLP::feedforward	(	const VectorFloat &	data,
		VectorFloat &	inputNeuronsOutput,
		VectorFloat &	hiddenNeuronsOutput,
		VectorFloat &	outputNeuronsOutput
	)

protected

Performs the feedforward step for back propagation, using the input data

Parameters

data	the input Vector to use for the feedforward
inputNeuronsOutput	the results of the input layer
hiddenNeuronsOutput	the results of the hidden layer
outputNeuronsOutput	the results of the output layer

Definition at line 966 of file MLP.cpp.

VectorFloat MLP::getClassDistances ( ) const

Gets a Vector of the class distances from the last prediction, this will be an N-dimensional Vector, where N is the number of classes in the model.

Returns: returns a Vector of the class distances from the last prediction, an empty Vector will be returned if the model has not been trained

Definition at line 1588 of file MLP.cpp.

bool MLP::getClassificationModeActive ( ) const

Returns true if the MLP is in classification mode.

Returns: returns true if the MLP is in classification mode, false otherwise

Definition at line 1542 of file MLP.cpp.

VectorFloat MLP::getClassLikelihoods ( ) const

Gets a Vector of the class likelihoods from the last prediction, this will be an N-dimensional Vector, where N is the number of classes in the model.

Returns: returns a Vector of the class likelihoods from the last prediction, an empty Vector will be returned if the model has not been trained

Definition at line 1583 of file MLP.cpp.

Float MLP::getGamma ( ) const

Gets the gamma value. This controls the gamma parameter for the neurons.

Returns: returns the gamma value for the neurons

Definition at line 1534 of file MLP.cpp.

Vector< Neuron > MLP::getHiddenLayer ( ) const

Returns the neurons for the hidden layer.

Returns: returns a Vector of neurons for the hidden layer

Definition at line 1554 of file MLP.cpp.

Neuron::Type MLP::getHiddenLayerActivationFunction ( ) const

Gets the hidden layer activation function.

Returns: returns the hidden layer activation function

Definition at line 1514 of file MLP.cpp.

std::string MLP::getId ( )

static

Gets a string that represents the MLP class.

Returns: returns a string containing the ID of this class

Definition at line 29 of file MLP.cpp.

Vector< Neuron > MLP::getInputLayer ( ) const

Returns the neurons for the input layer.

Returns: returns a Vector of neurons for the input layer

Definition at line 1550 of file MLP.cpp.

Neuron::Type MLP::getInputLayerActivationFunction ( ) const

Gets the input layer activation function.

Returns: returns the input layer activation function

Definition at line 1510 of file MLP.cpp.

Float MLP::getMaximumLikelihood ( ) const

Returns the current maximumLikelihood value. The maximumLikelihood value is computed during the prediction phase and is the likelihood of the most likely model. This is only relevant if the MLP is in classification mode. This value will return 0 if a prediction has not been made.

Returns: returns the current maximumLikelihood value

Definition at line 1578 of file MLP.cpp.

Float MLP::getMomentum ( ) const

Gets the momentum rate. This should be a value between [0 1]

Returns: returns the momentum

Definition at line 1530 of file MLP.cpp.

Float MLP::getNullRejectionCoeff ( ) const

Returns the current null rejection coefficient value. The null rejection coefficient is a multipler controlling the null rejection threshold for each class. This is only relevant if the MLP is in classification mode.

Returns: returns the current null rejection coefficient

Definition at line 1570 of file MLP.cpp.

bool MLP::getNullRejectionEnabled ( ) const

Returns true if the MLP should use null rejection during classification. This is only relevant if the MLP is in classification mode.

Returns: returns true if the null rejection is enabled, false otherwise

Definition at line 1566 of file MLP.cpp.

Float MLP::getNullRejectionThreshold ( ) const

Returns the current null rejection threshold value. The null rejection threshold is value controlling if a classification result should be rejected or accepted. This is only relevant if the MLP is in classification mode.

Returns: returns the current null rejection threshold

Definition at line 1574 of file MLP.cpp.

UINT MLP::getNumClasses ( ) const

Returns the number of classes in the MLP model if the MLP is in classification mode. The number of classes in the model is the same as the number of output neurons. If the MLP is in regression mode then it will return 0.

Returns: returns the number of classes in the MLP model if the MLP is in classification mode

Definition at line 1492 of file MLP.cpp.

UINT MLP::getNumHiddenNeurons ( ) const

Gets the number of hidden neurons.

Returns: returns the number of hidden neurons

Definition at line 1502 of file MLP.cpp.

UINT MLP::getNumInputNeurons ( ) const

Gets the number of input neurons.

Returns: returns the number of input neurons

Definition at line 1498 of file MLP.cpp.

UINT MLP::getNumOutputNeurons ( ) const

Gets the number of output neurons.

Returns: returns the number of output neurons

Definition at line 1506 of file MLP.cpp.

UINT MLP::getNumRandomTrainingIterations ( ) const

Gets the number of random training iterations that should be performed during the training phase. The MLP back propagation algorithm starts with random values, and the accuracy of a trained model can depend on which random values the algorithm started with. The GRT MLP algorithm therefore trains a number of models and picks the best one. This value therefore represents the number of random training iterations that should be used.

Returns: returns the number of random training iterations that should be performed during the training phase

Definition at line 1522 of file MLP.cpp.

Vector< Neuron > MLP::getOutputLayer ( ) const

Returns the neurons for the output layer.

Returns: returns a Vector of neurons for the output layer

Definition at line 1558 of file MLP.cpp.

Neuron::Type MLP::getOutputLayerActivationFunction ( ) const

Gets the output layer activation function.

Returns: returns the output layer activation function

Definition at line 1518 of file MLP.cpp.

UINT MLP::getPredictedClassLabel ( ) const

Gets the predicted class label from the last prediction. This is only relevant if the MLP is in classification mode.

Returns: returns the label of the last predicted class, a value of 0 will be returned if the model has not been trained

Definition at line 1594 of file MLP.cpp.

bool MLP::getRegressionModeActive ( ) const

Returns true if the MLP is in regression mode.

Returns: returns true if the MLP is in regression mode, false otherwise

Definition at line 1546 of file MLP.cpp.

Float MLP::getTrainingError ( ) const

Gets training error from the last round of training. If the MLP is in classification mode, the training error will be the classification accuracy. If the MLP is in regression mode then the training error will be the RMS error.

Returns: returns the training error from the last round of training

Definition at line 1538 of file MLP.cpp.

Vector< VectorFloat > MLP::getTrainingLog ( ) const

Returns a Vector of VectorFloat representing the training log for each random round of training. The outer Vector represents each round and the inner Vector represents each epoch in that round.

Returns: returns a Vector of MinMax values representing the ranges of the output layer

Definition at line 1562 of file MLP.cpp.

Float MLP::getTrainingRate ( ) const

Gets the training rate. This should be a value between [0 1]

Returns: returns the training rate

Definition at line 1526 of file MLP.cpp.

MLP::GRT_DEPRECATED_MSG	(	"setNumRandomTrainingIterations() is	deprecated,
		use setNumRestarts() instead"	,
		bool	setNumRandomTrainingIterationsconst UINT numRandomTrainingIterations
	)

Deprecated:: use setNumRestarts() instead Sets number of times the MLP model should be trained to find the best model. This value must be greater than zero.

Setting this value to a high number (i.e. 100) will most likely give you a better model, however it will take much longer to train the overall model. Setting this value to a low number (i.e. 5) will make the training process much faster, but you might not get the best model.

Parameters

numRandomTrainingIterations the number of times you want to randomly train the MLP model to search for the best results

Returns: returns true if the value was updated successfully, false otherwise

bool MLP::init	(	const UINT	numInputNeurons,
		const UINT	numHiddenNeurons,
		const UINT	numOutputNeurons
	)

Initializes the MLP for training. This should be called before the MLP is trained. The number of input neurons should match the number of input dimensions in your training data. The number of output neurons should match the number of target dimensions in your training data. The number of hidden units should be chosen by the user, a common rule of thumb is to set this as a value somewhere between the number of input neurons and the number of output neurons. Initializaling the MLP will clear any previous model or settings.

This function calls the other init function below, passing in the layer activation functions.

Parameters

numInputNeurons	the number of input neurons (should match the number of input dimensions in your training data)
numHiddenNeurons	the number of hidden neurons
numOutputNeurons	the number of output neurons (should match the number of target dimensions in your training data)

Returns: returns true if the MLP was initialized, false otherwise

Examples:: RegressionModulesExamples/MLPRegressionExample/MLPRegressionExample.cpp.

Definition at line 214 of file MLP.cpp.

bool MLP::init	(	const UINT	numInputNeurons,
		const UINT	numHiddenNeurons,
		const UINT	numOutputNeurons,
		const Neuron::Type	inputLayerActivationFunction,
		const Neuron::Type	hiddenLayerActivationFunction,
		const Neuron::Type	outputLayerActivationFunction
	)

Initializes the MLP for training. This should be called before the MLP is trained. The number of input neurons should match the number of input dimensions in your training data. The number of output neurons should match the number of target dimensions in your training data. The number of hidden units should be chosen by the user, a common rule of thumb is to set this as a value somewhere between the number of input neurons and the number of output neurons. The activation functions should be one of the Neuron ActivationFunctions enums. Initializaling the MLP will clear any previous model or settings.

Parameters

numInputNeurons	the number of input neurons (should match the number of input dimensions in your training data)
numHiddenNeurons	the number of hidden neurons
numOutputNeurons	the number of output neurons (should match the number of target dimensions in your training data)
inputLayerActivationFunction	the activation function to use for the input layer
hiddenLayerActivationFunction	the activation function to use for the input layer
outputLayerActivationFunction	the activation function to use for the input layer

Returns: returns true if the MLP was initialized, false otherwise

Definition at line 218 of file MLP.cpp.

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

virtual

This loads a trained MLP model from a file. This overrides the load function in the ML base class.

Parameters

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

Returns: returns true if the model was loaded successfully, false otherwise

Reimplemented from MLBase.

Definition at line 1148 of file MLP.cpp.

MLP & MLP::operator= ( const MLP & rhs )

Defines how the data from the rhs MLP should be copied to this MLP

Parameters

rhs	another instance of a MLP

Returns: returns a reference to this instance of the MLP

Definition at line 67 of file MLP.cpp.

bool MLP::predict_ ( VectorFloat & inputVector )

virtual

This function either predicts the class of the input Vector (if the MLP is in Classification Mode), or it performs regression using the MLP model.

Parameters

inputVector the input Vector to classify or perform regression on

Returns: returns true if the prediction/regression was performed, false otherwise

Reimplemented from MLBase.

Definition at line 155 of file MLP.cpp.

bool MLP::print ( ) const

virtual

This function will print the model and settings to the display log.

Returns: returns true if the model was printed succesfully, false otherwise

Reimplemented from MLBase.

Definition at line 307 of file MLP.cpp.

void MLP::printNetwork ( ) const

Prints the current MLP weights and coefficents to std out. This function is depreciated, you should now use print() instead.

Definition at line 993 of file MLP.cpp.

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

virtual

This saves the trained MLP model to a file. This overrides the save function in the ML base class.

Parameters

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

Returns: returns true if the model was saved successfully, false otherwise

Reimplemented from MLBase.

Definition at line 1074 of file MLP.cpp.

bool MLP::setGamma ( const Float gamma )

Sets the gamma parameter for the Neurons. Gamma must be greater than zero. If the MLP instance has been initialized then this function will also call the init function to reinitialize the instance.

Parameters

gamma the gamma value for each Neuron, gamma must be greater than zero

Returns: returns true if the value was updated successfully, false otherwise

Definition at line 1709 of file MLP.cpp.

bool MLP::setHiddenLayerActivationFunction ( const Neuron::Type activationFunction )

This function sets the activation function for all the Neurons in the hidden layer. If the MLP instance has been initialized then this function will also call the init function to reinitialize the instance.

Parameters

activationFunction the activation function for the hidden layer, this should be one of the Neuron ActivationFunctions enums

Returns: returns true if the hidden layer activation function was set successfully, false otherwise

Definition at line 1666 of file MLP.cpp.

bool MLP::setInputLayerActivationFunction ( const Neuron::Type activationFunction )

This function sets the activation function for all the Neurons in the input layer. If the MLP instance has been initialized then this function will also call the init function to reinitialize the instance.

Parameters

activationFunction the activation function for the input layer, this should be one of the Neuron ActivationFunctions enums

Returns: returns true if the input layer activation function was set successfully, false otherwise

Definition at line 1650 of file MLP.cpp.

bool MLP::setMomentum ( const Float momentum )

Sets the momentum parameter. This is used to update the weights at each step of the stochastic gradient descent. The momentum parameter is normally set between [0.1 0.9], with 0.5 being a common value.

Parameters

momentum the momentum value used during the training phase, must be greater than zero

Returns: returns true if the value was updated successfully, false otherwise

Examples:: RegressionModulesExamples/MLPRegressionExample/MLPRegressionExample.cpp.

Definition at line 1701 of file MLP.cpp.

bool MLP::setNullRejection ( const bool useNullRejection )

Sets if null rejection should be used for the real-time prediction. This is only used if the MLP is in classificationMode.

Parameters

useNullRejection if true then null rejection will be used

Returns: returns true if the value was updated successfully, false otherwise

Definition at line 1728 of file MLP.cpp.

bool MLP::setNullRejectionCoeff ( const Float nullRejectionCoeff )

This function lets you manually control the null rejection threshold. Any class with a prediction value less than the null rejection threshold will be rejected, setting the predicted class label to 0.

This is only used if the MLP is in classificationMode.

Parameters

nullRejectionCoeff the new null rejection threshold

Returns: returns true if the value was updated successfully, false otherwise

Definition at line 1733 of file MLP.cpp.

bool MLP::setOutputLayerActivationFunction ( const Neuron::Type activationFunction )

This function sets the activation function for all the Neurons in the output layer. If the MLP instance has been initialized then this function will also call the init function to reinitialize the instance.

Parameters

activationFunction the activation function for the output layer, this should be one of the Neuron ActivationFunctions enums

Returns: returns true if the output layer activation function was set successfully, false otherwise

Definition at line 1682 of file MLP.cpp.

bool MLP::setTrainingRate ( const Float trainingRate )

Sets the training rate, which controls the learning rate parameter. This is used to update the weights at each step of the stochastic gradient descent. The learningRate value must be greater than zero, a value of 0.1 normally works well. If you find the MLP fails to train with this value then try setting the learning rate to a smaller value (for example 0.01).

Parameters

trainingRate the learningRate value used during the training phase, must be greater than zero

Returns: returns true if the value was updated successfully, false otherwise

Definition at line 1697 of file MLP.cpp.

bool MLP::train_ ( ClassificationData & trainingData )

virtual

This trains the MLP model, using the labelled classification data. Calling this function sets the MLP into Classification Model.

Parameters

trainingData the training data that will be used to train the classification model

Returns: returns true if the MLP model was trained, false otherwise

Reimplemented from MLBase.

Definition at line 121 of file MLP.cpp.

bool MLP::train_ ( RegressionData & trainingData )

virtual

This trains the MLP model, using the labelled regression data. Calling this function sets the MLP into Regression Model.

Parameters

trainingData the training data that will be used to train the regression model

Returns: returns true if the MLP model was trained, false otherwise

Reimplemented from MLBase.

Definition at line 146 of file MLP.cpp.

bool MLP::validateActivationFunction ( const Neuron::Type avactivationFunction ) const

Validates if the activationFunction is valid.

Parameters

activationFunction the activation function you want to valid

Returns: returns true if the activation function is valid, false otherwise

Definition at line 1645 of file MLP.cpp.

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

Public Types

Public Member Functions

Static Public Member Functions

Protected Member Functions

Protected Attributes

Additional Inherited Members

Detailed Description

Constructor & Destructor Documentation

Member Function Documentation