Evonet Class Reference

Evonet is the neural network taken from the EvoRobot. More...

Inheritance diagram for Evonet:

Signals
void	evonetUpdated ()
	emitted everytime the Evonet has been updated

Public Member Functions
	Evonet ()
	The class constructor.
float	backPropStep (QVector< float > tInput, double rate)
	Computes the neuron deltas for the Backpropagation algorithm.
float	backPropStep2 (QVector< float > tInput, double rate)
	Computes the neuron deltas for the Backpropagation algorithm but does not update the weights.
void	calculateBackPropagationError (QVector< float > tInput)
	Computes the Backpropagation error (the backpropagation algorithm is not applied to the ANN)
float	computeMeanSquaredError (QVector< float > trainingSet, QVector< float > desiredOutput)
	Update the network and compute and return the mean squared error.
void	computeParameters ()
	Compute the required number of free parameters on the basis of: (i) the property of the neurons (stored in the vectors neurontype[], neuronbias[], neurongain[]) (ii) the connection blocks (stored in the matrix net_block[b][0])
void	configure (ConfigurationParameters &params, QString prefix)
	Configures the object using a ConfigurationParameters object.
void	copyPheParameters (int *pheGene)
	transorm floating point parameters normalized in the range [-wrange,range] into integer parameters in the range [0,255]
void	create_net_block (int inputNeuronType, int hiddenNeuronType, int outputNeuronType, bool recurrentHiddens, bool inputOutputConnections, bool recurrentOutputs, bool biasOnHidden, bool biasOnOutput)
	Create the block structure that describe the architecture of the network (unless this structure is created through the graphic interface and loaded froa a .net or .phe file) Update blocks describe group of neurons to be updated Connections blocks describe the group of neurons receiving the connections and the group sending the connections Gain blocks describe a group of neurons that share the same gain of that have a gain regulated by a third neuron This functions also set whether neurons have biases, timeconstant, and gains the neuron update function (e.g.
void	endTraining ()
	de-allocate variables create for training
int	exportWeightsToMATLABFile (char *path)
int	extractWeightsFromNet (Eigen::VectorXf &w)
	Extract the connection weights and biases from the free parameters vector into a matrix to be used by the Levenberg Marquardt Algoririthm.
int	freeParameters ()
	return number of defined free parameters
float	getBackPropError ()
	Returns the Backpropagation algorithm error.
float	getBrange ()
	return the value of brange
float	getFreeParameter (int i)
	return the value of the ith parameter (normalized in the range [-wrange, wrange]
float	getGrange ()
	return the value of grange
float	getHidden (int h)
	return the value of a hidden neuron (-999 if the specified id is out of range)
float	getInput (int in)
	return the value of a sensory neuron
void	getMutations (float *mut)
	set the mutation vector of the genetic algorithm that determine how parameters are mutated
const QString &	getNetworkName () const
	Returns the name of this neural network.
float	getNeuron (int in)
	return the activation of a neuron
int	getNoHiddens ()
	return the number of internal neurons
int	getNoInputs ()
	return the number of sensory neurons
int	getNoNeurons ()
	return the total number of neurons
int	getNoOutputs ()
	return the number of motor neurons
float *	getOldestStoredActivations ()
	Returns the oldest stored activation vector and removes it.
float	getOutput (int out)
	return the value of a motor neuron (-1 if the specified id is out of range)
int	getParamBias (int nbias)
	return the value of a bias
float	getTeachingInputEntry (int id)
	Returns the value of the teaching input element in id position.
ParameterSettableUI *	getUIManager ()
	Return a new instance of the EvonetUI.
float	getWeight (int to, int from)
	Return the weight that depart from neuron "from" and reach neuron "to".
float	getWrange ()
	return the value of wrange (which also determine the range in which all parameters are normalized)
void	hardwire ()
	Experiment specific function that set the weights for a reflex and set the weights that should not be trained to 0.0.
int	importWeightsFromMATLABFile (char *path)
int	importWeightsFromVector (Eigen::VectorXf &w)
	Copy back the weights and biases trained through the Levenberg Marquardt Algorithm into the free parameters vector.
void	initBackPropFreep ()
	Initializes backpropfreep for applying backPropStep2 algorithm.
void	initWeightsInRange (float min, float max)
	Initializes weights and biases randomly.
void	initWeightsInRange (float minBias, float maxBias, float minWeight, float maxWeight)
	Initializes weights and biases randomly.
void	initWeightsNguyenWidrow (float min, float max)
	Initializes weights and biases according to the Nguyen-Widrow initialization algorithm.
void	injectHidden (int nh, float val)
	set the actiovation state of a hidden neuron
bool	inTraining ()
int	load_net_blocks (const char *filename, int mode)
	Load the description of the neural architecture from .net of .phe file (see the create_net_block method description)
float	logistic (float f)
	logistic function
bool	pheFileLoaded ()
	check whether a .phe file (with parameters description) has been loaded
void	prepareForTraining (QVector< float > &err_w)
	Initialize variables required by back-propagation training mconnections: number of weights and biases subject to training outputsToTrain: outputsToTrain[i]==1 iff the i-th output is subject to training n_outputsToTrain: number of output neurons to train trainingHiddenBlock: a matrix of noutputs value for each neuron.
void	printAct ()
	Print neurons activation state.
void	printBlocks ()
	display the architecture structure
void	printIO ()
	Print the activation state of sensory, internal, and motor neurons.
void	printWeights ()
	Print all free parameters.
void	readNewPheLine (QStringList, float *, float *)
	Assign to a free parameter and to the free parameter mutation rate the value extracted from a string loaded from a file .phe file.
void	readOldPheLine (QStringList, float *, float *)
	Assign to a free parameter and to the free parameter mutation rate the value extracted from a string loaded from a file .phe file This function is used instead of the readNewPheLine() method to parse .phe files that contain a single QstringList element.
void	resetNet ()
	reset to 0.0 the activation state of all neurons
void	save (ConfigurationParameters &params, QString prefix)
	Saves the actual status of parameters into the ConfigurationParameters object passed.
void	save_net_blocks (const char *filename, int mode)
	Save the description of the neural architecture into a .net of .phe file (see the create_net_block method description)
int	setInput (int inp, float value)
	set the value of a sensory neuron
void	setNeckReflex ()
	Experiment specific function that set certain weights so to provide a given reflex behavior.
void	setNetworkName (const QString &name)
	Sets the name of this neural network.
void	setParameters (const int *dt)
	set the free parameters on the basis of a genotype string
void	setRanges (double weight, double bias, double gain)
	set the range of connection weights, biases, and gains
void	setWeight (int to, int from, float w)
	Set the weight that depart from neuron "from" and reach neuron "to".
bool	showTeachingInput ()
	Checks whether the teaching input has to be shown.
float	tansig (float f)
	Hyperbolic tangent sigmoid transfer function.
float	trainLevembergMarquardt (QVector< float > trainingSet, QVector< float > desiredOutput, float maxError)
	Train the network through the Levemberg Marquardt gradient descent algorithm.
float	trainLevembergMarquardtThroughTime (QVector< float > trainingSet, QVector< float > desiredOutput, int time, float maxError)
	Train the network through the Levemberg Marquardt through time gradient descent algorithm.
int	updateCounts ()
	return the number of updates (step) done
void	updateNet ()
	Update the state of the internal and motor neurons on the basis of: (i) the property of the neurons (stored in the vectors neurontype[], neuronbias[], neurongain[]) (ii) the update, connection, and gain blocks (stored in the matrix net_block[b][0]) (iii) the value of the free parameters.
void	updateWeightsAfterBackProp ()
	Updates the weights of the ANN.
Public Member Functions inherited from ParameterSettableWithConfigureFunction
void	addObserver (RuntimeParameterObserver *obs)
T	getRuntimeParameter (QString paramName)
	ParameterSettable ()
virtual void	postConfigureInitialization ()
void	removeObserver (RuntimeParameterObserver *obs)
void	setRuntimeParameter (QString paramName, T newvalue)
QString	typeName () const

Static Public Member Functions
static void	describe (QString type)
	Add to Factory::typeDescriptions() the descriptions of all parameters and subgroups.
Static Public Member Functions inherited from ParameterSettableWithConfigureFunction
static QString	fullParameterDescriptionPath (QString type, QString param)
static QString	fullSubgroupDescriptionPath (QString type, QString sub)

Public Attributes
int	maxIterations
	Maximum number of training iteration.
QColor	neurondcolor [MAXN]
	the color used to display the actiovation state of each neuron in the neuron monitor widget
int	neurondisplay [MAXN]
	the vectors that specify for each neuron whether it should be displayed or not by the neuron monitor widget
char	neuronl [MAXN][10]
	the labels of the neurons displayed by the graphic widget
bool	neuronlesion [MAXN]
	a vector that speficy lesioned and unlesioned neurons
int	neuronlesions
	Whether one or more neurons have been lesioned.
float	neuronlesionVal [MAXN]
	the value to be assigned to the state of lesioned neurons
double	neuronrange [MAXN][2]
	the matrix that contain the variation range of neurons used by the neuron monitor graphic widget
int	nselected
	Number of parameters selected through the graphic interface Enable the user to modify the parameters manually through the graphic widget.
float **	selectedp
	pointer to the list pointer-to-parameters selected through the graphic interface Enable the user to modify the parameters manually through the graphic widget
Public Attributes inherited from ParameterSettableWithConfigureFunction
	AllowMultiple
	Default
	IsList
	IsMandatory
Public Attributes inherited from ParameterSettable
	AllowMultiple
	Default
	IsList
	IsMandatory

Static Public Attributes
static const float	DEFAULT_VALUE = -99.0f
	DEFAULT_VALUE is used for do not assign values to mut and parameters.
static const int	MAXN = 1000
	Maximum number of neurons for a neural network of this type.
static const int	MAXSTOREDACTIVATIONS = 100
	The maximum number of stored activation vectors.
Static Public Attributes inherited from ParameterSettableWithConfigureFunction
static const double	Infinity
static const int	MaxInteger
static const int	MinInteger
Static Public Attributes inherited from ParameterSettable
static const double	Infinity
static const int	MaxInteger
static const int	MinInteger

Friends
class	NetworkDialog
class	RendNetwork

Additional Inherited Members
Public Types inherited from ParameterSettableWithConfigureFunction
enum	Property
Protected Member Functions inherited from ParameterSettableWithConfigureFunction
void	notifyChangesToParam (QString paramName)
Static Protected Member Functions inherited from ParameterSettableWithConfigureFunction
static Descriptor	addTypeDescription (QString type, QString shortHelp, QString longHelp=QString(""))
static void	setGraphicalEditor (QString type)

Detailed Description

Evonet is the neural network taken from the EvoRobot.

Simple class to load a evorobot.net configuration file

Definition at line 100 of file evonet.h.

Constructor & Destructor Documentation

Evonet

(

)

The class constructor.

Definition at line 62 of file evonet.cpp.

References Evonet::MAXN, Evonet::neurondcolor, Evonet::neurondisplay, Evonet::neuronl, Evonet::neuronlesion, Evonet::neuronlesions, Evonet::neuronlesionVal, Evonet::neuronrange, and Evonet::selectedp.

Member Function Documentation

float backPropStep	(	QVector< float >	tInput,
		double	rate
	)

Computes the neuron deltas for the Backpropagation algorithm.

Parameters

tInput	the input vector
rate	the delta rate

Returns

the Backpropagation error

Definition at line 1916 of file evonet.cpp.

References DataUploader< DataType_t >::getDownloader(), and Evonet::MAXN.

float backPropStep2	(	QVector< float >	tInput,
		double	rate
	)

Computes the neuron deltas for the Backpropagation algorithm but does not update the weights.

Parameters

tInput	the input vector
rate	the delta rate

Returns

the Backpropagation error

Definition at line 2030 of file evonet.cpp.

References DataUploader< DataType_t >::getDownloader(), and Evonet::MAXN.

void calculateBackPropagationError

(

QVector< float >

tInput

)

Computes the Backpropagation error (the backpropagation algorithm is not applied to the ANN)

Parameters

tInput

the input vector

Definition at line 2145 of file evonet.cpp.

float computeMeanSquaredError	(	QVector< float >	trainingSet,
		QVector< float >	desiredOutput
	)

Update the network and compute and return the mean squared error.

Parameters

trainingSet	the input vector
desiredOutput	the teaching vector

Definition at line 1470 of file evonet.cpp.

References Evonet::setInput(), and Evonet::updateNet().

Referenced by Evonet::trainLevembergMarquardt(), and Evonet::trainLevembergMarquardtThroughTime().

void computeParameters

(

)

Compute the required number of free parameters on the basis of: (i) the property of the neurons (stored in the vectors neurontype[], neuronbias[], neurongain[]) (ii) the connection blocks (stored in the matrix net_block[b][0])

Definition at line 695 of file evonet.cpp.

References Evonet::MAXN, and Logger::warning().

Referenced by Evonet::configure(), Evonet::create_net_block(), and Evonet::save_net_blocks().

void configure ( ConfigurationParameters &  params, QString  prefix  )

virtual

Configures the object using a ConfigurationParameters object.

Parameters

params	the configuration parameters object with parameters to use
prefix	the prefix to use to access the object configuration parameters. This is guaranteed to end with the separator character when called by the factory, so you don't need to add one

Implements ParameterSettableWithConfigureFunction.

Definition at line 105 of file evonet.cpp.

References Evonet::computeParameters(), Evonet::create_net_block(), Evonet::DEFAULT_VALUE, Logger::error(), ConfigurationHelper::getBool(), ConfigurationHelper::getDouble(), ConfigurationHelper::getInt(), ConfigurationHelper::getString(), Evonet::load_net_blocks(), Evonet::MAXN, Evonet::neurondcolor, Evonet::neuronl, Evonet::neuronrange, Evonet::printBlocks(), Evonet::resetNet(), and ConfigurationHelper::throwUserConfigError().

void copyPheParameters

(

int *

pheGene

)

transorm floating point parameters normalized in the range [-wrange,range] into integer parameters in the range [0,255]

Parameters

pheGene

the vector of integers to be filled with the current parameters

Definition at line 989 of file evonet.cpp.

References Evonet::DEFAULT_VALUE, and Evonet::freeParameters().

Referenced by TestRandom::buildRandomDNA(), and Evoga::getPheParametersAndMutationsFromEvonet().

void create_net_block	(	int	inputNeuronType,
		int	hiddenNeuronType,
		int	outputNeuronType,
		bool	recurrentHiddens,
		bool	inputOutputConnections,
		bool	recurrentOutputs,
		bool	biasOnHidden,
		bool	biasOnOutput
	)

Create the block structure that describe the architecture of the network (unless this structure is created through the graphic interface and loaded froa a .net or .phe file) Update blocks describe group of neurons to be updated Connections blocks describe the group of neurons receiving the connections and the group sending the connections Gain blocks describe a group of neurons that share the same gain of that have a gain regulated by a third neuron This functions also set whether neurons have biases, timeconstant, and gains the neuron update function (e.g.

logistic, leaky, binary) and the x,y positions in which the neurons are visualized in the graphic widgets

Definition at line 266 of file evonet.cpp.

References Evonet::computeParameters(), and Evonet::neurondisplay.

Referenced by Evonet::configure().

void describe ( QString  type )

static

Add to Factory::typeDescriptions() the descriptions of all parameters and subgroups.

Reimplemented from ParameterSettableWithConfigureFunction.

Definition at line 242 of file evonet.cpp.

References ParameterSettableWithConfigureFunction::addTypeDescription(), ParameterSettable::RealDescriptor::def(), ParameterSettable::BoolDescriptor::def(), ParameterSettable::EnumDescriptor::def(), ParameterSettable::Descriptor::describeBool(), ParameterSettable::Descriptor::describeEnum(), ParameterSettable::Descriptor::describeInt(), ParameterSettable::Descriptor::describeReal(), ParameterSettable::Descriptor::describeString(), ParameterSettable::StringDescriptor::help(), ParameterSettable::BoolDescriptor::help(), ParameterSettable::IntDescriptor::help(), ParameterSettable::RealDescriptor::help(), ParameterSettable::EnumDescriptor::help(), ParameterSettableWithConfigureFunction::Infinity, ParameterSettable::RealDescriptor::limits(), ParameterSettable::IntDescriptor::limits(), Evonet::MAXN, and ParameterSettable::EnumDescriptor::values().

void endTraining

(

)

de-allocate variables create for training

Definition at line 1892 of file evonet.cpp.

void evonetUpdated ( )

signal

emitted everytime the Evonet has been updated

Note

connect this signal to all necessary slots for updating the graphics stuff

Referenced by Evonet::setParameters(), and Evonet::updateNet().

int extractWeightsFromNet

(

Eigen::VectorXf &

w

)

Extract the connection weights and biases from the free parameters vector into a matrix to be used by the Levenberg Marquardt Algoririthm.

Definition at line 1514 of file evonet.cpp.

Referenced by Evonet::trainLevembergMarquardt(), and Evonet::trainLevembergMarquardtThroughTime().

int freeParameters

(

)

return number of defined free parameters

Definition at line 951 of file evonet.cpp.

Referenced by Evonet::copyPheParameters(), Evonet::getMutations(), Evoga::getPheParametersAndMutationsFromEvonet(), Evonet::getWeight(), Evonet::initWeightsInRange(), Evonet::printWeights(), Evonet::save_net_blocks(), and Evonet::setParameters().

float getBackPropError

(

)

Returns the Backpropagation algorithm error.

Returns

the backpropagation error

Definition at line 1911 of file evonet.cpp.

float getBrange

(

)

return the value of brange

Definition at line 1036 of file evonet.cpp.

float getFreeParameter

(

int

i

)

return the value of the ith parameter (normalized in the range [-wrange, wrange]

Definition at line 956 of file evonet.cpp.

Referenced by Evoga::updateGenomeFromEvonet().

float getGrange

(

)

return the value of grange

Definition at line 1041 of file evonet.cpp.

float getHidden

(

int

h

)

return the value of a hidden neuron (-999 if the specified id is out of range)

Parameters

h	the id number of the motor neuron

Definition at line 942 of file evonet.cpp.

float getInput

(

int

in

)

return the value of a sensory neuron

Parameters

in	the id number of the motor neuron

Definition at line 914 of file evonet.cpp.

Referenced by EvonetIterator::getInput().

void getMutations

(

float *

mut

)

set the mutation vector of the genetic algorithm that determine how parameters are mutated

Parameters

mut	the pointer to the evonet mutation vector loaded from a.phe file

Definition at line 981 of file evonet.cpp.

References Evonet::freeParameters().

Referenced by Evoga::getPheParametersAndMutationsFromEvonet().

const QString & getNetworkName

(

)

const

Returns the name of this neural network.

Returns

the name of this neural network

Definition at line 100 of file evonet.cpp.

float getNeuron

(

int

in

)

return the activation of a neuron

Parameters

in	the id number of the neuron

Definition at line 919 of file evonet.cpp.

int getNoHiddens

(

)

return the number of internal neurons

Definition at line 1070 of file evonet.cpp.

Referenced by EvonetIterator::setGraphicProperties().

int getNoInputs

(

)

return the number of sensory neurons

Definition at line 1065 of file evonet.cpp.

Referenced by EvonetIterator::setGraphicProperties().

int getNoNeurons

(

)

return the total number of neurons

Definition at line 1080 of file evonet.cpp.

int getNoOutputs

(

)

return the number of motor neurons

Definition at line 1075 of file evonet.cpp.

float * getOldestStoredActivations

(

)

Returns the oldest stored activation vector and removes it.

This returns the oldest activation vector and deletes it (memory is not freed, simply an internal index is incremented), so that subsequent calls will progressively return newer activation vectors. Returns NULL when no activation vector is stored

Returns

the oldest stored activation or NULL if no stored activation vector is present

Definition at line 1092 of file evonet.cpp.

References Evonet::MAXSTOREDACTIVATIONS.

float getOutput

(

int

out

)

return the value of a motor neuron (-1 if the specified id is out of range)

Parameters

out	the id number of the motor neuron

Definition at line 906 of file evonet.cpp.

Referenced by EvonetIterator::getOutput().

int getParamBias

(

int

nbias

)

return the value of a bias

Parameters

h	the id number of the neuron

Definition at line 1022 of file evonet.cpp.

float getTeachingInputEntry

(

int

id

)

Returns the value of the teaching input element in id position.

Parameters

id	the index of the teaching input desired value

Returns

the value of the teaching input element

Definition at line 1906 of file evonet.cpp.

ParameterSettableUI * getUIManager ( )

virtual

Return a new instance of the EvonetUI.

Reimplemented from ParameterSettableWithConfigureFunction.

Definition at line 262 of file evonet.cpp.

float getWeight	(	int	to,
		int	from
	)

Return the weight that depart from neuron "from" and reach neuron "to".

Definition at line 1762 of file evonet.cpp.

References Evonet::freeParameters().

Referenced by Evonet::initWeightsNguyenWidrow(), Evonet::trainLevembergMarquardt(), and Evonet::trainLevembergMarquardtThroughTime().

float getWrange

(

)

return the value of wrange (which also determine the range in which all parameters are normalized)

Definition at line 1031 of file evonet.cpp.

Referenced by Evoga::updateGenomeFromEvonet().

void hardwire

(

)

Experiment specific function that set the weights for a reflex and set the weights that should not be trained to 0.0.

Definition at line 1224 of file evonet.cpp.

int importWeightsFromVector

(

Eigen::VectorXf &

w

)

Copy back the weights and biases trained through the Levenberg Marquardt Algorithm into the free parameters vector.

Definition at line 1643 of file evonet.cpp.

Referenced by Evonet::trainLevembergMarquardt(), and Evonet::trainLevembergMarquardtThroughTime().

void initBackPropFreep

(

)

Initializes backpropfreep for applying backPropStep2 algorithm.

Definition at line 2162 of file evonet.cpp.

void initWeightsInRange	(	float	min,
		float	max
	)

Initializes weights and biases randomly.

Parameters

min	the minimum value of the weights
max	the maximum value of the weights

Definition at line 1116 of file evonet.cpp.

References Evonet::freeParameters(), and farsa::min().

Referenced by Evonet::initWeightsNguyenWidrow().

void initWeightsInRange	(	float	minBias,
		float	maxBias,
		float	minWeight,
		float	maxWeight
	)

Initializes weights and biases randomly.

Here you can specify different ranges for biases and weights

 \param minBias the minimum value of the biases
 \param maxBias the maximum value of the biases
 \param minWeight the minimum value of the weights
 \param maxWeight the maximum value of the weights

Definition at line 1142 of file evonet.cpp.

References Evonet::freeParameters().

void initWeightsNguyenWidrow	(	float	min,
		float	max
	)

Initializes weights and biases according to the Nguyen-Widrow initialization algorithm.

Parameters

min	the minimum value of the weights
max	the maximum value of the weights

Definition at line 1170 of file evonet.cpp.

References Evonet::getWeight(), Evonet::initWeightsInRange(), and Evonet::setWeight().

void injectHidden	(	int	nh,
		float	val
	)

set the actiovation state of a hidden neuron

Parameters

nh	the id number of the hidden
val	the activation value to be set

Definition at line 935 of file evonet.cpp.

int load_net_blocks	(	const char *	filename,
		int	mode
	)

Load the description of the neural architecture from .net of .phe file (see the create_net_block method description)

Parameters

filename	the filename in which the description is contained
mode	0=load only the architecture (.net file) 1=load also the value of the parameters (.phe file)

Definition at line 433 of file evonet.cpp.

References Logger::error(), Logger::info(), Evonet::MAXN, Evonet::neurondisplay, Evonet::readNewPheLine(), Evonet::readOldPheLine(), and Logger::warning().

Referenced by Evonet::configure().

float logistic

(

float

f

)

logistic function

Parameters

f	the input value of the function

Definition at line 682 of file evonet.cpp.

Referenced by Evonet::updateNet().

bool pheFileLoaded

(

)

check whether a .phe file (with parameters description) has been loaded

Definition at line 961 of file evonet.cpp.

Referenced by Evoga::getPheParametersAndMutationsFromEvonet().

void prepareForTraining

(

QVector< float > &

err_w

)

Initialize variables required by back-propagation training mconnections: number of weights and biases subject to training outputsToTrain: outputsToTrain[i]==1 iff the i-th output is subject to training n_outputsToTrain: number of output neurons to train trainingHiddenBlock: a matrix of noutputs value for each neuron.

Parameters

err_w

???

Definition at line 1814 of file evonet.cpp.

void printAct

(

)

Print neurons activation state.

Definition at line 1461 of file evonet.cpp.

void printBlocks

(

)

display the architecture structure

Definition at line 1047 of file evonet.cpp.

References Logger::info().

Referenced by Evonet::configure().

void printIO

(

)

Print the activation state of sensory, internal, and motor neurons.

Definition at line 1001 of file evonet.cpp.

References Logger::info().

void printWeights

(

)

Print all free parameters.

Definition at line 1108 of file evonet.cpp.

References Evonet::freeParameters().

void readNewPheLine	(	QStringList	line,
		float *	par,
		float *	mut
	)

Assign to a free parameter and to the free parameter mutation rate the value extracted from a string loaded from a file .phe file.

Parameters

QStringList	the string from which the parameter is extracted, the first element contain the value or an "*" which means that the value of the parameter is initialized randomly the second element contain a mutation rate or "*" which means that the parameter has the default mutation rate
float*	the pointer to the free parameter
float*	the pointer to the mutation rate of the free parameter

Definition at line 539 of file evonet.cpp.

References Evonet::DEFAULT_VALUE.

Referenced by Evonet::load_net_blocks().

void readOldPheLine	(	QStringList	line,
		float *	par,
		float *	mut
	)

Assign to a free parameter and to the free parameter mutation rate the value extracted from a string loaded from a file .phe file This function is used instead of the readNewPheLine() method to parse .phe files that contain a single QstringList element.

Parameters

QStringList	the string from which the parameter is extracted, the first element contain the value or an "*" which means that the value of the parameter is initialized randomly since the second element is missing, the parameter is considered as fixed (i.e. the mutation rate of the parameter is set to 0.0)
float*	the pointer to the free parameter
float*	the pointer to the mutation rate of the free parameter

Definition at line 530 of file evonet.cpp.

References Evonet::DEFAULT_VALUE.

Referenced by Evonet::load_net_blocks().

void resetNet

(

)

reset to 0.0 the activation state of all neurons

Definition at line 924 of file evonet.cpp.

References Evonet::MAXN.

Referenced by Evonet::configure().

void save ( ConfigurationParameters &  params, QString  prefix  )

virtual

Saves the actual status of parameters into the ConfigurationParameters object passed.

Parameters

params	the configuration parameters object on which save actual parameters
prefix	the prefix to use to access the object configuration parameters.

Implements ParameterSettableWithConfigureFunction.

Definition at line 228 of file evonet.cpp.

References ConfigurationParameters::createParameter(), and ConfigurationParameters::startObjectParameters().

void save_net_blocks	(	const char *	filename,
		int	mode
	)

Save the description of the neural architecture into a .net of .phe file (see the create_net_block method description)

Parameters

filename	the filename in which the description is contained
mode	0=save only the architecture (.net file) 1=save also the value of the parameters (.phe file)

Definition at line 576 of file evonet.cpp.

References Evonet::computeParameters(), Evonet::DEFAULT_VALUE, Logger::error(), Evonet::freeParameters(), Logger::info(), Evonet::neurondisplay, and Evonet::neuronl.

int setInput	(	int	inp,
		float	value
	)

set the value of a sensory neuron

Parameters

inp	the id number of the sensory neuron
value	the value to be set

Definition at line 897 of file evonet.cpp.

Referenced by Evonet::computeMeanSquaredError(), EvonetIterator::setInput(), Evonet::trainLevembergMarquardt(), and Evonet::trainLevembergMarquardtThroughTime().

void setNeckReflex

(

)

Experiment specific function that set certain weights so to provide a given reflex behavior.

Definition at line 2972 of file evonet.cpp.

void setNetworkName

(

const QString &

name

)

Sets the name of this neural network.

This is just stored and not used anywhere in this class.

Parameters

name	the name of the network

Definition at line 95 of file evonet.cpp.

void setParameters

(

const int *

dt

)

set the free parameters on the basis of a genotype string

Parameters

df	the pointer to the vector of genotype integers set in the range [0, 255]

Definition at line 969 of file evonet.cpp.

References Evonet::evonetUpdated(), and Evonet::freeParameters().

void setRanges	(	double	weight,
		double	bias,
		double	gain
	)

set the range of connection weights, biases, and gains

Parameters

weight	the range of weights
bias	the range of biases
gain	the range of gains

Definition at line 1085 of file evonet.cpp.

void setWeight	(	int	to,
		int	from,
		float	w
	)

Set the weight that depart from neuron "from" and reach neuron "to".

Definition at line 1787 of file evonet.cpp.

Referenced by Evonet::initWeightsNguyenWidrow().

bool showTeachingInput

(

)

Checks whether the teaching input has to be shown.

Returns

TRUE if the teaching input must be shown, FALSE otherwise

Definition at line 1900 of file evonet.cpp.

float tansig

(

float

f

)

Hyperbolic tangent sigmoid transfer function.

Definition at line 687 of file evonet.cpp.

float trainLevembergMarquardt	(	QVector< float >	trainingSet,
		QVector< float >	desiredOutput,
		float	maxError
	)

Train the network through the Levemberg Marquardt gradient descent algorithm.

Parameters

trainingSet	the input vector
desiredOutput	the teaching vector
maxError	the desired error threshold

Definition at line 2202 of file evonet.cpp.

References Evonet::computeMeanSquaredError(), Evonet::extractWeightsFromNet(), Evonet::getWeight(), Evonet::importWeightsFromVector(), farsa::max(), Evonet::maxIterations, Evonet::setInput(), and Evonet::updateNet().

float trainLevembergMarquardtThroughTime	(	QVector< float >	trainingSet,
		QVector< float >	desiredOutput,
		int	time,
		float	maxError
	)

Train the network through the Levemberg Marquardt through time gradient descent algorithm.

Parameters

trainingSet	the input vector
desiredOutput	the teaching vector
time	the number of interations in time
maxError	the desired error threshold

Definition at line 2512 of file evonet.cpp.

References Evonet::computeMeanSquaredError(), Evonet::extractWeightsFromNet(), Evonet::getWeight(), Evonet::importWeightsFromVector(), farsa::max(), Evonet::maxIterations, Evonet::setInput(), and Evonet::updateNet().

int updateCounts

(

)

return the number of updates (step) done

The counter is resetted to zero when the Evonet is resetted (Evonet::resetNet) and is incremented by one at each call of Evonet::updateNet

Definition at line 1103 of file evonet.cpp.

void updateNet

(

)

Update the state of the internal and motor neurons on the basis of: (i) the property of the neurons (stored in the vectors neurontype[], neuronbias[], neurongain[]) (ii) the update, connection, and gain blocks (stored in the matrix net_block[b][0]) (iii) the value of the free parameters.

Definition at line 752 of file evonet.cpp.

References Evonet::evonetUpdated(), DataUploader< DataType_t >::getDownloader(), Evonet::logistic(), Evonet::MAXN, Evonet::MAXSTOREDACTIVATIONS, Evonet::neuronlesion, Evonet::neuronlesions, and Evonet::neuronlesionVal.

Referenced by Evonet::computeMeanSquaredError(), Evonet::trainLevembergMarquardt(), and Evonet::trainLevembergMarquardtThroughTime().

void updateWeightsAfterBackProp

(

)

Updates the weights of the ANN.

Definition at line 2175 of file evonet.cpp.

Member Data Documentation

const float DEFAULT_VALUE = -99.0f

static

DEFAULT_VALUE is used for do not assign values to mut and parameters.

Definition at line 116 of file evonet.h.

Referenced by TestRandom::buildRandomDNA(), Evonet::configure(), Evoga::copyGenes(), Evonet::copyPheParameters(), Evoga::getGenome(), Evoga::postConfigureInitialization(), Evonet::readNewPheLine(), Evonet::readOldPheLine(), Evonet::save_net_blocks(), and Evoga::setInitialPopulation().

int maxIterations

Maximum number of training iteration.

Definition at line 570 of file evonet.h.

Referenced by Evonet::trainLevembergMarquardt(), and Evonet::trainLevembergMarquardtThroughTime().

const int MAXN = 1000

static

Maximum number of neurons for a neural network of this type.

Definition at line 114 of file evonet.h.

Referenced by Evonet::backPropStep(), Evonet::backPropStep2(), Evonet::computeParameters(), Evonet::configure(), Evonet::describe(), Evonet::Evonet(), Evonet::load_net_blocks(), Evonet::resetNet(), and Evonet::updateNet().

const int MAXSTOREDACTIVATIONS = 100

static

The maximum number of stored activation vectors.

Definition at line 112 of file evonet.h.

Referenced by Evonet::getOldestStoredActivations(), and Evonet::updateNet().

QColor neurondcolor[MAXN]

the color used to display the actiovation state of each neuron in the neuron monitor widget

Definition at line 361 of file evonet.h.

Referenced by Evonet::configure(), Evonet::Evonet(), and EvonetIterator::setGraphicProperties().

int neurondisplay[MAXN]

the vectors that specify for each neuron whether it should be displayed or not by the neuron monitor widget

Definition at line 352 of file evonet.h.

Referenced by Evonet::create_net_block(), Evonet::Evonet(), Evonet::load_net_blocks(), and Evonet::save_net_blocks().

char neuronl[MAXN][10]

the labels of the neurons displayed by the graphic widget

Definition at line 348 of file evonet.h.

Referenced by Evonet::configure(), Evonet::Evonet(), Evonet::save_net_blocks(), and EvonetIterator::setGraphicProperties().

bool neuronlesion[MAXN]

a vector that speficy lesioned and unlesioned neurons

Definition at line 365 of file evonet.h.

Referenced by Evonet::Evonet(), and Evonet::updateNet().

int neuronlesions

Whether one or more neurons have been lesioned.

Definition at line 409 of file evonet.h.

Referenced by Evonet::Evonet(), and Evonet::updateNet().

float neuronlesionVal[MAXN]

the value to be assigned to the state of lesioned neurons

Definition at line 369 of file evonet.h.

Referenced by Evonet::Evonet(), and Evonet::updateNet().

double neuronrange[MAXN][2]

the matrix that contain the variation range of neurons used by the neuron monitor graphic widget

Definition at line 357 of file evonet.h.

Referenced by Evonet::configure(), Evonet::Evonet(), and EvonetIterator::setGraphicProperties().

int nselected

Number of parameters selected through the graphic interface Enable the user to modify the parameters manually through the graphic widget.

Definition at line 405 of file evonet.h.

float** selectedp

pointer to the list pointer-to-parameters selected through the graphic interface Enable the user to modify the parameters manually through the graphic widget

Definition at line 400 of file evonet.h.

Referenced by Evonet::Evonet().

---

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

• experiments/include/evonet.h
• experiments/src/evonet.cpp