Contains Parameters that define neural network architecture and trainig procedure, together with achieved results after training such as various error norms. More...

Collaboration diagram for IG.Num.NeuralTrainingParameters:

Classes
class	ComparerBase
	Base comparer class (implementation of the IComparer{NeuralTrainingParameters} interface) for conmparing objects of type NeuralTrainingParameters More...

class	ComparerErrorTrainingVector

class	ComparerTrainingTime

Public Member Functions
void	SetEpochNumbers (int[] epochNumbers)
	Sets the list of epoch numbers at which convergence data was sampled. More...

void	SetEpochNumbers (List< int > epochNumbers)
	Sets the list of epoch numbers at which convergence data was sampled. More...

void	SetEpochErrorsRms (double[] errors)
	Sets the list of sampled RMS errors that correspond to epoch numbers from EpochNumbers. More...

void	SetEpochErrorsRms (List< double > errors)
	Sets the list of sampled RMS errors corresponding to epoch numbers from EpochNumbers. More...

void	SetEpochErrorsAbs (double[] errors)
	Sets the list of sampled absolute errors that correspond to epoch numbers from EpochNumbers. More...

void	SetEpochErrorsAbs (List< double > errors)
	Sets the list of sampled absolute errors corresponding to epoch numbers from EpochNumbers. More...

void	CopyFrom (NeuralApproximatorBase nn)
	Copies current data from the specified neural network approximator. More...

void	CopyResultsFrom (NeuralApproximatorBase nn)
	Copies only results from the trained network to the current object. More...

void	CopyTo (NeuralApproximatorBase nn)
	Copies data that determine neural network and training procedure (such as network architecture, training parameters, tolerances, etc.) from the current object to the specified neural network approximator. This enables to restore training contitions of a peviour training procedure. More...

Static Public Member Functions
static void	SaveJson (NeuralTrainingParameters trainingParameters, string filePath)
	Saves (serializes) the specified training parameters object to the specified JSON file. File is owerwritten if it exists. More...

static void	SaveJson (NeuralTrainingParameters trainingParameters, string filePath, bool append)
	Saves (serializes) the specified training parameters object to the specified JSON file. More...

static void	LoadJson (string filePath, ref NeuralTrainingParameters trainingParametersRestored)
	Restores (deserializes) a training parameters object from the specified file in JSON format. More...

static void	SaveJson (NeuralTrainingParameters[] trainingParameters, string filePath)
	Saves (serializes) the specified array of training parameters objects to the specified JSON file. File is owerwritten if it exists. More...

static void	SaveJson (NeuralTrainingParameters[] trainingParameters, string filePath, bool append)
	Saves (serializes) the specified array of training parameters objects to the specified JSON file. More...

static void	LoadJson (string filePath, ref NeuralTrainingParameters[] trainingParametersRestored)
	Restores (deserializes) an array of training parameters objects from the specified file in JSON format. More...

static void	SaveJson (List< NeuralTrainingParameters > trainingParameters, string filePath)
	Saves (serializes) the specified list of training parameters objects to the specified JSON file. File is owerwritten if it exists. More...

static void	SaveJson (List< NeuralTrainingParameters > trainingParameters, string filePath, bool append)
	Saves (serializes) the specified list of training parameters objects to the specified JSON file. More...

static void	LoadJson (string filePath, ref List< NeuralTrainingParameters > trainingParametersRestored)
	Restores (deserializes) a list of training parameters objects from the specified file in JSON format. More...

static void	SaveCSV (List< NeuralTrainingParameters > trainingParameters, string filePath)
	Saves the specified list of training parameters objects to the specified CSV file. More...

static ComparerTrainingTime	CreateComparerTrainingTime ()

Public Attributes
List< IVector >	_errorsTrainingRmsList = null

List< IVector >	_errorsTrainingMaxList = null

List< IVector >	_errorsVerificationRmsList = null

List< IVector >	_errorsVerificationMaxList = null

bool	_convergenceRmsEnabled = false

Static Public Attributes
static double	DefaultLearningRate = 0.1
	Default value for learning rate in neural networks. More...

static double	DefaultMomentum = 0.5
	Default value of momentum for neural networks. More...

static double	DefaultSigmoidAlphaValue = 1.5
	Default value of the sigmoid alpha value (used in networks with sigmoid activation functions). More...

static int	DefaultEpochsInBundle
	Default value for number of epochs in bundle (i.e. number of epochs performed at once, without any checks or output operations between). More...

static int	DefaultNumHiddenLayers = 1
	Default number of hidden layers. More...

static int	DefaultNumHiddenNeurons = 20
	Default number of hidden neurons in a layer More...

static int	DefaultMaxEpochs = 40000
	Default value for maximal number of epochs. More...

static IVector	DefaultToleranceRms = null
	Default value for tollerance on RMS error in neural networks. More...

static IVector	DefaultToleranceMax = null
	Default value for tollerance on max. abs. error in neural networks. More...

static double	DefaultToleranceRmsRelativeToRangeScalar = 0.1
	Default value for the tolerance on RMS error, relative to the output range. More...

static double	DefaultToleranceMaxRelativeToRangeScalar = 0
	Default value for the tolerance on max. abs. error, relative to the output range. More...

static int	DefaultInputLength = 1
	Default number of input neurons. More...

static int	DefaultOutputLength = 1
	Default number of output neurons. More...

static double	DefaultInputBoundSafetyFactor = 1.5
	Default number of input safety factor. More...

static double	DefaultOutputBoundSafetyFactor = 1.5
	Default number of output safety factor. More...

Protected Member Functions
IComparer< IVector >	CreateComparerErrorTrainingRmsVector ()

Static Protected Member Functions
static double	ErrorAverageCalc (List< IVector > ErrorConvergence, int NumLastErrors, int NumBundles, IVector scalingLength)
	Calculate average number of the desired number of first elements in the convergence list. More...

static double	ErrorAverageCalc (List< IVector > ErrorConvergence, int NumLastErrors, int NumBundles)
	Calculate average number of the desired number of first elements in the convergence list. More...

Protected Attributes
double	_learningRate = DefaultLearningRate

double	_momentum = DefaultMomentum

double	_sigmoidAlphaValue = DefaultSigmoidAlphaValue

int	_maxEpochs = DefaultMaxEpochs

int	_epochInBundle = DefaultEpochsInBundle

IVector	_outputRange

IBoundingBox	_outputNeuronRange

IVector	_inputRange

IBoundingBox	_inputNeuronRange

IVector	_toleranceRms

IVector	_toleranceMax

int	_numHiddenLayers = DefaultNumHiddenLayers

int[]	_numHiddenNeurons = null

bool	_isNetworkTrained = false

IVector	_errorsTrainingRms

IVector	_errorsTrainingMax

IVector	_errorsTrainingMeanAbs

IVector	_errorsVerificationRms

IVector	_errorsVerificationMax

IVector	_errorsVerificationMeanAbs

int	_numEpochs = 0

double	_trainingTime = 0

double	_trainingCpuTime = 0

List< int >	_EpochNumbers

List< double >	_EpochErrorsRms

List< double >	_EpochErrorsAbs

Properties
double	LearningRate `[get, set]`
	Learning rate. More...

double	Momentum `[get, set]`
	Momentum. Specifies how much changes of weight in the previous iterations affect changes in the current iterations. More...

double	SigmoidAlphaValue `[get, set]`
	Sigmoid alpha value (used in networks with sigmoid activation functions). More...

virtual int	InputLength `[get, set]`
	Gets or sets the number of input neurons. More...

virtual int	OutputLength `[get, set]`
	Gets or sets the number of output neurons. More...

virtual double	InputBoundSafetyFactor `[get, set]`
	Gets or sets input safety factor. More...

virtual double	OutputBoundSafetyFactor `[get, set]`
	Gets or sets output safety factor. More...

int	MaxEpochs `[get, set]`
	Maximal number of epochs performed in the training procedure. More...

int	EpochsInBundle `[get, set]`
	Number of epochs in bundle (i.e. number of epochs performed at once, without any checks or output operations between). More...

IVector	OutputRange `[get, set]`
	Range from actual outputs. More...

IBoundingBox	OutputNeuronRange `[get, set]`
	Bounding box from actual outputs. More...

IVector	InputRange `[get, set]`
	Range from actual inputs. More...

IBoundingBox	InputNeuronRange `[get, set]`
	Bounding box from actual inputs. More...

IVector	ToleranceRms `[get, set]`
	Tolerance over RMS error of output over training points. Training will continue until error becomes below tolerance or until maximal number of epochs is reached. If less or equal than 0 then this tolerance is not taken into account. More...

IVector	ToleranceMax `[get, set]`
	Tolerance on maximal error of output over training points. Training will continue until error becomes below tolerance or until maximal number of epochs is reached. If less or equal than 0 then this tolerance is not taken into account. More...

IVector	ToleranceRmsRelativeToRange `[get, set]`
	Relative tolerances on RMS errors of outputs over training points, relative to the correspoinding ranges of output data. More...

double	ToleranceRmsRelativeToRangeScalar `[get, set]`
	Scalar through which all components of the Relative tolerances on RMS errors of outputs can be set to the same value. More...

IVector	ToleranceMaxRelativeToRange `[get, set]`
	Relative tolerances on max. abs. errors of outputs over training points, relative to the correspoinding ranges of output data. More...

double	ToleranceMaxRelativeToRangeScalar `[get, set]`
	Scalar through which all components of the Relative tolerances on max. abs. errors of outputs can be set to the same value. More...

int	NumHiddenLayers `[get, set]`
	Number of Hideden layers. More...

int[]	NumHidenNeurons `[get, set]`
	Numbers of neurons in each hidden layer. More...

bool	IsNetworkTrained `[get, set]`
	Whether the network is trained (and results exist). More...

IVector	ErrorsTrainingRms `[get, set]`
	RMS errors calculated on training data. More...

List< IVector >	ErrorsTrainingRmsList `[get, set]`
	Convergence List of Rms errors calculated on training data. More...

IVector	ErrorsTrainingMax `[get, set]`
	Maximal errors calculated on training data. More...

List< IVector >	ErrorsTrainingMaxList `[get, set]`
	Convergence List of Maximal errors calculated on training data. More...

IVector	ErrorsTrainingMeanAbs `[get, set]`
	Mean absolute errors calculated on training data. More...

IVector	ErrorsVerificationRms `[get, set]`
	RMS errors calculated on verification data. More...

List< IVector >	ErrorsVerificationRmsList `[get, set]`
	Convergence List of RMS errors calculated on verification data. More...

IVector	ErrorsVerificationMax `[get, set]`
	Maximal errors calculated on verification data. More...

List< IVector >	ErrorsVerificationMaxList `[get, set]`
	Convergence List of Maximal errors calculated on verification data. More...

IVector	ErrorsVerificationMeanAbs `[get, set]`
	Maximal errors calculated on training data. More...

bool	SaveConvergenceRms `[get, set]`

int	NumEpochs `[get, set]`
	Number of epochs actually spent at training. More...

double	TrainingTime `[get, set]`
	Time spent for training. More...

double	TrainingCpuTime `[get, set]`
	CPU time spent for training. More...

List< int >	EpochNumbers `[get, protected set]`
	List of epoch numbers at which convergence data was sampled. More...

List< double >	EpochErrorsRms `[get, set]`
	List of sampled RMS errors corresponding to epoch numbers from EpochNumbers. More...

List< double >	EpochErrorsAbs `[get, set]`
	List of sampled absolute errors corresponding to epoch numbers from EpochNumbers. More...

Private Attributes
int	_inputLength = DefaultInputLength

int	_outputLength = DefaultOutputLength

double	_inputBoundSafetyFactor = DefaultInputBoundSafetyFactor
	Gets or sets input safety factor. More...

double	_outputBoundSafetyFactor = DefaultOutputBoundSafetyFactor
	Gets or sets output safety factor. More...

IVector	_tolRmsRelative
	Auxiliary properties for defining tolerances in a relative way: More...

double	_tolRmsRelativeScalar = DefaultToleranceRmsRelativeToRangeScalar

IVector	_tolMaxRelative

double	_tolMaxRelativeScalar = DefaultToleranceMaxRelativeToRangeScalar

Detailed Description

Contains Parameters that define neural network architecture and trainig procedure, together with achieved results after training such as various error norms.

Not thread safe!

This class is used for storing parameters of neural networks and restoring them at a later time, in order to repeat training under similar condition or simply to analyse performance of neural networks.

$A Igor Jul10 May12;

Member Function Documentation

void IG.Num.NeuralTrainingParameters.SetEpochNumbers ( int[] epochNumbers )

inline

Sets the list of epoch numbers at which convergence data was sampled.

Parameters

epochNumbers Array of epoch numbers from which data is copied.

void IG.Num.NeuralTrainingParameters.SetEpochNumbers ( List< int > epochNumbers )

inline

Sets the list of epoch numbers at which convergence data was sampled.

Parameters

epochNumbers List of epoch numbers from which data is copied.

void IG.Num.NeuralTrainingParameters.SetEpochErrorsRms ( double[] errors )

inline

Sets the list of sampled RMS errors that correspond to epoch numbers from EpochNumbers.

Parameters

errors Array from which data is copied.

void IG.Num.NeuralTrainingParameters.SetEpochErrorsRms ( List< double > errors )

inline

Sets the list of sampled RMS errors corresponding to epoch numbers from EpochNumbers.

Parameters

errors List from which data is copied.

void IG.Num.NeuralTrainingParameters.SetEpochErrorsAbs ( double[] errors )

inline

Sets the list of sampled absolute errors that correspond to epoch numbers from EpochNumbers.

Parameters

errors Array from which data is copied.

void IG.Num.NeuralTrainingParameters.SetEpochErrorsAbs ( List< double > errors )

inline

Sets the list of sampled absolute errors corresponding to epoch numbers from EpochNumbers.

Parameters

errors List from which data is copied.

void IG.Num.NeuralTrainingParameters.CopyFrom ( NeuralApproximatorBase nn )

inline

Copies current data from the specified neural network approximator.

Parameters

nn	Neural network approximator that data is copied form.

void IG.Num.NeuralTrainingParameters.CopyResultsFrom ( NeuralApproximatorBase nn )

inline

Copies only results from the trained network to the current object.

Parameters

nn	Neural network approximator which results are copied from.

void IG.Num.NeuralTrainingParameters.CopyTo ( NeuralApproximatorBase nn )

inline

Copies data that determine neural network and training procedure (such as network architecture, training parameters, tolerances, etc.) from the current object to the specified neural network approximator. This enables to restore training contitions of a peviour training procedure.

Parameters

nn	Neural network approximator that data is copied to.

static void IG.Num.NeuralTrainingParameters.SaveJson	(	NeuralTrainingParameters	trainingParameters,
		string	filePath
	)

inlinestatic

Saves (serializes) the specified training parameters object to the specified JSON file. File is owerwritten if it exists.

Parameters

trainingParameters	Object that is saved to a file.
filePath	Path to the file into which object is is saved.

static void IG.Num.NeuralTrainingParameters.SaveJson	(	NeuralTrainingParameters	trainingParameters,
		string	filePath,
		bool	append
	)

inlinestatic

Saves (serializes) the specified training parameters object to the specified JSON file.

Parameters

trainingParameters	Object that is saved to a file.
filePath	Path to the file into which object is is saved.
append	Specifies whether serialized data is appended at the end of the file in the case that the file already exists.

static void IG.Num.NeuralTrainingParameters.LoadJson	(	string	filePath,
		ref NeuralTrainingParameters	trainingParametersRestored
	)

inlinestatic

Restores (deserializes) a training parameters object from the specified file in JSON format.

Parameters

filePath	File from which object data is restored.
trainingParametersRestored	Object that is restored by deserialization.

static void IG.Num.NeuralTrainingParameters.SaveJson	(	NeuralTrainingParameters[]	trainingParameters,
		string	filePath
	)

inlinestatic

Saves (serializes) the specified array of training parameters objects to the specified JSON file. File is owerwritten if it exists.

Parameters

trainingParameters	Array that is saved to a file.
filePath	Path to the file into which object is is saved.

static void IG.Num.NeuralTrainingParameters.SaveJson	(	NeuralTrainingParameters[]	trainingParameters,
		string	filePath,
		bool	append
	)

inlinestatic

Saves (serializes) the specified array of training parameters objects to the specified JSON file.

Parameters

trainingParameters	Array that is saved to a file.
filePath	Path to the file into which object is is saved.
append	Specifies whether serialized data is appended at the end of the file in the case that the file already exists.

static void IG.Num.NeuralTrainingParameters.LoadJson	(	string	filePath,
		ref NeuralTrainingParameters[]	trainingParametersRestored
	)

inlinestatic

Restores (deserializes) an array of training parameters objects from the specified file in JSON format.

Parameters

filePath	File from which array of objects is restored.
trainingParametersRestored	Array of objects that is restored by deserialization.

static void IG.Num.NeuralTrainingParameters.SaveJson	(	List< NeuralTrainingParameters >	trainingParameters,
		string	filePath
	)

inlinestatic

Saves (serializes) the specified list of training parameters objects to the specified JSON file. File is owerwritten if it exists.

Parameters

trainingParameters	List that is saved to a file.
filePath	Path to the file into which object is is saved.

static void IG.Num.NeuralTrainingParameters.SaveJson	(	List< NeuralTrainingParameters >	trainingParameters,
		string	filePath,
		bool	append
	)

inlinestatic

Saves (serializes) the specified list of training parameters objects to the specified JSON file.

Parameters

trainingParameters	List that is saved to a file.
filePath	Path to the file into which object is is saved.
append	Specifies whether serialized data is appended at the end of the file in the case that the file already exists.

static void IG.Num.NeuralTrainingParameters.LoadJson	(	string	filePath,
		ref List< NeuralTrainingParameters >	trainingParametersRestored
	)

inlinestatic

Restores (deserializes) a list of training parameters objects from the specified file in JSON format.

Parameters

filePath	File from which list of objects is restored.
trainingParametersRestored	Array of objects that is restored by deserialization.

static void IG.Num.NeuralTrainingParameters.SaveCSV	(	List< NeuralTrainingParameters >	trainingParameters,
		string	filePath
	)

inlinestatic

Saves the specified list of training parameters objects to the specified CSV file.

Parameters

trainingParameters	List that is saved to a file.
filePath	Path to the file into which object is is saved.

$A Tako78 Sep12;

static double IG.Num.NeuralTrainingParameters.ErrorAverageCalc	(	List< IVector >	ErrorConvergence,
		int	NumLastErrors,
		int	NumBundles,
		IVector	scalingLength
	)

inlinestaticprotected

Calculate average number of the desired number of first elements in the convergence list.

Parameters

ErrorConvergence	List of convergence errors.
NumLastErrors	Number of elemnts in the convergence list.
NumBundles	Number of epoch bundles.
scalingLength	Scaling length.

Returns: Average number.

References IG.Num.VectorBase.NormWeighted().

static double IG.Num.NeuralTrainingParameters.ErrorAverageCalc	(	List< IVector >	ErrorConvergence,
		int	NumLastErrors,
		int	NumBundles
	)

inlinestaticprotected

Calculate average number of the desired number of first elements in the convergence list.

Parameters

ErrorConvergence	List of convergence errors.
NumLastErrors	Number of elemnts in the convergence list.
NumBundles	Number of bundles.

Returns: Average number.

IComparer<IVector> IG.Num.NeuralTrainingParameters.CreateComparerErrorTrainingRmsVector ( )

inlineprotected

static ComparerTrainingTime IG.Num.NeuralTrainingParameters.CreateComparerTrainingTime ( )

inlinestatic

Member Data Documentation

double IG.Num.NeuralTrainingParameters.DefaultLearningRate = 0.1

static

Default value for learning rate in neural networks.

double IG.Num.NeuralTrainingParameters.DefaultMomentum = 0.5

static

Default value of momentum for neural networks.

double IG.Num.NeuralTrainingParameters.DefaultSigmoidAlphaValue = 1.5

static

Default value of the sigmoid alpha value (used in networks with sigmoid activation functions).

int IG.Num.NeuralTrainingParameters.DefaultEpochsInBundle

static

Default value for number of epochs in bundle (i.e. number of epochs performed at once, without any checks or output operations between).

int IG.Num.NeuralTrainingParameters.DefaultNumHiddenLayers = 1

static

Default number of hidden layers.

int IG.Num.NeuralTrainingParameters.DefaultNumHiddenNeurons = 20

static

Default number of hidden neurons in a layer

int IG.Num.NeuralTrainingParameters.DefaultMaxEpochs = 40000

static

Default value for maximal number of epochs.

IVector IG.Num.NeuralTrainingParameters.DefaultToleranceRms = null

static

Default value for tollerance on RMS error in neural networks.

IVector IG.Num.NeuralTrainingParameters.DefaultToleranceMax = null

static

Default value for tollerance on max. abs. error in neural networks.

double IG.Num.NeuralTrainingParameters.DefaultToleranceRmsRelativeToRangeScalar = 0.1

static

Default value for the tolerance on RMS error, relative to the output range.

double IG.Num.NeuralTrainingParameters.DefaultToleranceMaxRelativeToRangeScalar = 0

static

Default value for the tolerance on max. abs. error, relative to the output range.

int IG.Num.NeuralTrainingParameters.DefaultInputLength = 1

static

Default number of input neurons.

int IG.Num.NeuralTrainingParameters.DefaultOutputLength = 1

static

Default number of output neurons.

double IG.Num.NeuralTrainingParameters.DefaultInputBoundSafetyFactor = 1.5

static

Default number of input safety factor.

double IG.Num.NeuralTrainingParameters.DefaultOutputBoundSafetyFactor = 1.5

static

Default number of output safety factor.

double IG.Num.NeuralTrainingParameters._learningRate = DefaultLearningRate

protected

double IG.Num.NeuralTrainingParameters._momentum = DefaultMomentum

protected

double IG.Num.NeuralTrainingParameters._sigmoidAlphaValue = DefaultSigmoidAlphaValue

protected

int IG.Num.NeuralTrainingParameters._inputLength = DefaultInputLength

private

int IG.Num.NeuralTrainingParameters._outputLength = DefaultOutputLength

private

double IG.Num.NeuralTrainingParameters._inputBoundSafetyFactor = DefaultInputBoundSafetyFactor

private

Gets or sets input safety factor.

double IG.Num.NeuralTrainingParameters._outputBoundSafetyFactor = DefaultOutputBoundSafetyFactor

private

Gets or sets output safety factor.

int IG.Num.NeuralTrainingParameters._maxEpochs = DefaultMaxEpochs

protected

int IG.Num.NeuralTrainingParameters._epochInBundle = DefaultEpochsInBundle

protected

IVector IG.Num.NeuralTrainingParameters._outputRange

protected

IBoundingBox IG.Num.NeuralTrainingParameters._outputNeuronRange

protected

IVector IG.Num.NeuralTrainingParameters._inputRange

protected

IBoundingBox IG.Num.NeuralTrainingParameters._inputNeuronRange

protected

IVector IG.Num.NeuralTrainingParameters._toleranceRms

protected

IVector IG.Num.NeuralTrainingParameters._toleranceMax

protected

IVector IG.Num.NeuralTrainingParameters._tolRmsRelative

private

Auxiliary properties for defining tolerances in a relative way:

double IG.Num.NeuralTrainingParameters._tolRmsRelativeScalar = DefaultToleranceRmsRelativeToRangeScalar

private

IVector IG.Num.NeuralTrainingParameters._tolMaxRelative

private

double IG.Num.NeuralTrainingParameters._tolMaxRelativeScalar = DefaultToleranceMaxRelativeToRangeScalar

private

int IG.Num.NeuralTrainingParameters._numHiddenLayers = DefaultNumHiddenLayers

protected

int [] IG.Num.NeuralTrainingParameters._numHiddenNeurons = null

protected

bool IG.Num.NeuralTrainingParameters._isNetworkTrained = false

protected

IVector IG.Num.NeuralTrainingParameters._errorsTrainingRms

protected

List<IVector> IG.Num.NeuralTrainingParameters._errorsTrainingRmsList = null

IVector IG.Num.NeuralTrainingParameters._errorsTrainingMax

protected

List<IVector> IG.Num.NeuralTrainingParameters._errorsTrainingMaxList = null

IVector IG.Num.NeuralTrainingParameters._errorsTrainingMeanAbs

protected

IVector IG.Num.NeuralTrainingParameters._errorsVerificationRms

protected

List<IVector> IG.Num.NeuralTrainingParameters._errorsVerificationRmsList = null

IVector IG.Num.NeuralTrainingParameters._errorsVerificationMax

protected

List<IVector> IG.Num.NeuralTrainingParameters._errorsVerificationMaxList = null

IVector IG.Num.NeuralTrainingParameters._errorsVerificationMeanAbs

protected

bool IG.Num.NeuralTrainingParameters._convergenceRmsEnabled = false

int IG.Num.NeuralTrainingParameters._numEpochs = 0

protected

double IG.Num.NeuralTrainingParameters._trainingTime = 0

protected

double IG.Num.NeuralTrainingParameters._trainingCpuTime = 0

protected

List<int> IG.Num.NeuralTrainingParameters._EpochNumbers

protected

List<double> IG.Num.NeuralTrainingParameters._EpochErrorsRms

protected

List<double> IG.Num.NeuralTrainingParameters._EpochErrorsAbs

protected

Property Documentation

double IG.Num.NeuralTrainingParameters.LearningRate

getset

Learning rate.

Referenced by IG.Num.NeuralTrainingParametersDto.CopyFromPlain(), IG.Script.LoadableScriptShellNeuralBase.Custom(), and IG.Num.NeuralTrainingLimits.IterateThroughMultidimensionalTable().

double IG.Num.NeuralTrainingParameters.Momentum

getset

Momentum. Specifies how much changes of weight in the previous iterations affect changes in the current iterations.

Referenced by IG.Num.NeuralTrainingParametersDto.CopyFromPlain(), and IG.Num.NeuralTrainingLimits.IterateThroughMultidimensionalTable().

double IG.Num.NeuralTrainingParameters.SigmoidAlphaValue

getset

Sigmoid alpha value (used in networks with sigmoid activation functions).

Referenced by IG.Num.NeuralTrainingParametersDto.CopyFromPlain(), and IG.Num.NeuralTrainingLimits.IterateThroughMultidimensionalTable().

virtual int IG.Num.NeuralTrainingParameters.InputLength

getset

Gets or sets the number of input neurons.

Referenced by IG.Num.NeuralTrainingParametersDto.CopyFromPlain(), and IG.Num.NeuralTrainingLimits.IterateThroughMultidimensionalTable().

virtual int IG.Num.NeuralTrainingParameters.OutputLength

getset

Gets or sets the number of output neurons.

Referenced by IG.Num.NeuralTrainingParametersDto.CopyFromPlain(), and IG.Num.NeuralTrainingLimits.IterateThroughMultidimensionalTable().

virtual double IG.Num.NeuralTrainingParameters.InputBoundSafetyFactor

getset

Gets or sets input safety factor.

Referenced by IG.Num.NeuralTrainingParametersDto.CopyFromPlain(), and IG.Num.NeuralTrainingLimits.IterateThroughMultidimensionalTable().

virtual double IG.Num.NeuralTrainingParameters.OutputBoundSafetyFactor

getset

Gets or sets output safety factor.

Referenced by IG.Num.NeuralTrainingParametersDto.CopyFromPlain(), and IG.Num.NeuralTrainingLimits.IterateThroughMultidimensionalTable().

int IG.Num.NeuralTrainingParameters.MaxEpochs

getset

Maximal number of epochs performed in the training procedure.

Referenced by IG.Num.NeuralTrainingParametersDto.CopyFromPlain(), and IG.Num.NeuralTrainingLimits.IterateThroughMultidimensionalTable().

int IG.Num.NeuralTrainingParameters.EpochsInBundle

getset

Number of epochs in bundle (i.e. number of epochs performed at once, without any checks or output operations between).

This parameter does not affect the training procedure in terms of results.

Referenced by IG.Num.NeuralTrainingParametersDto.CopyFromPlain(), and IG.Num.NeuralTrainingLimits.IterateThroughMultidimensionalTable().

IVector IG.Num.NeuralTrainingParameters.OutputRange

getset

Range from actual outputs.

$A Tako78 Octl12;

Referenced by IG.Num.NeuralTrainingParametersDto.CopyFromPlain().

IBoundingBox IG.Num.NeuralTrainingParameters.OutputNeuronRange

getsetprotected

Bounding box from actual outputs.

$A Tako78 Octl12;

IVector IG.Num.NeuralTrainingParameters.InputRange

getset

Range from actual inputs.

$A Tako78 Octl12;

Referenced by IG.Num.NeuralTrainingParametersDto.CopyFromPlain().

IBoundingBox IG.Num.NeuralTrainingParameters.InputNeuronRange

getsetprotected

Bounding box from actual inputs.

$A Tako78 Octl12;

IVector IG.Num.NeuralTrainingParameters.ToleranceRms

getset

Tolerance over RMS error of output over training points. Training will continue until error becomes below tolerance or until maximal number of epochs is reached. If less or equal than 0 then this tolerance is not taken into account.

Referenced by IG.Num.NeuralTrainingParametersDto.CopyFromPlain(), and IG.Num.NeuralTrainingLimits.IterateThroughMultidimensionalTable().

IVector IG.Num.NeuralTrainingParameters.ToleranceMax

getset

Tolerance on maximal error of output over training points. Training will continue until error becomes below tolerance or until maximal number of epochs is reached. If less or equal than 0 then this tolerance is not taken into account.

Referenced by IG.Num.NeuralTrainingParametersDto.CopyFromPlain(), and IG.Num.NeuralTrainingLimits.IterateThroughMultidimensionalTable().

IVector IG.Num.NeuralTrainingParameters.ToleranceRmsRelativeToRange

getset

Relative tolerances on RMS errors of outputs over training points, relative to the correspoinding ranges of output data.