MatrixLinker Class define a full connection between two cluster. More...

Inheritance diagram for MatrixLinker:

Public Member Functions
	MatrixLinker (Cluster from, Cluster to, QString name="unnamed")
	Connect clusters with a complete connections By default it create a fully-connected matrix, use one of the following methods for choose a different way: connect, connectRandom, disconnect, disconnectRandom.
	MatrixLinker (ConfigurationParameters &params, QString prefix)
	Constructor.
virtual	~MatrixLinker ()
	Destructor.
unsigned int	cols ()
	Get the number of cols.
void	connect (unsigned int from, unsigned int to)
	Connect two neurons.
void	connectAll ()
	Connect all couples of neurons.
void	connectRandom (double prob, bool zeroDiagonal=false, bool symmetric=false)
	Connect neurons of Clusters with a random connections with the passed probability.
void	disconnect (unsigned int from, unsigned int to)
	Disconnect the two neurons.
void	disconnectAll ()
	Disonnect all couples of neurons.
DoubleMatrix	matrix () const
	const version of matrix() method
DoubleMatrix &	matrix ()
	Return the weight matrix.
virtual void	randomize (double min, double max)
	Randomize the weights of the MatrixLinker.
unsigned int	rows ()
	Get the number of rows.
virtual void	save (ConfigurationParameters &params, QString prefix)
	Save the actual status of parameters into the ConfigurationParameters object passed.
virtual void	setWeight (unsigned int from, unsigned int to, double weight)
	Set the weight of the connection specified.
unsigned int	size () const
	Returns the total number of the links: rows*cols.
virtual double	weight (unsigned int from, unsigned int to)
	Get the weight of the connection specified.

Detailed Description

MatrixLinker Class define a full connection between two cluster.

Motivation: This class rapresent connections between neuron's 'from' Cluster and neuron's 'to' Cluster with a matrix. It allow sparse connections as full connections amogns neurons. The connections has a weight and are stored in a matrix which the rows are the neuron's 'from' and the columns the neuron's 'to'.

Description: Every connection is weighted, and the weight is memorized into a weight-matrix.
The effective computation of inputs' 'to' is done in the subclasses (DotLinker, NormLinker, etc). The disconnection / connection of neurons are implemented fixating/unfixating the weight to zero into the underlying matrix using DoubleMatrix::steady and DoubleMatrix::unsteady methods

Warning

Definition at line 47 of file matrixlinker.h.

Constructor & Destructor Documentation

MatrixLinker	(	Cluster *	from,
		Cluster *	to,
		QString	name = `"unnamed"`
	)

Connect clusters with a complete connections By default it create a fully-connected matrix, use one of the following methods for choose a different way: connect, connectRandom, disconnect, disconnectRandom.

Definition at line 25 of file matrixlinker.cpp.

References DoubleMatrix::zeroing().

MatrixLinker	(	ConfigurationParameters &	params,
		QString	prefix
	)

Constructor.

Definition at line 30 of file matrixlinker.cpp.

References ConfigurationParameters::getValue().

~MatrixLinker ( ) [virtual]

Destructor.

Definition at line 56 of file matrixlinker.cpp.

Member Function Documentation

unsigned int cols ( ) [inline]

Get the number of cols.

Definition at line 63 of file matrixlinker.h.

Referenced by MatrixLinker::connectAll(), MatrixLinker::connectRandom(), MatrixLinker::disconnectAll(), MatrixLinker::save(), and NormLinker::update().

void connect	(	unsigned int	from,
		unsigned int	to
	)

Connect two neurons.

Definition at line 79 of file matrixlinker.cpp.

References DoubleMatrix::unsteady().

void connectAll ( )

Connect all couples of neurons.

Definition at line 130 of file matrixlinker.cpp.

References MatrixLinker::cols(), MatrixLinker::rows(), and DoubleMatrix::unsteady().

void connectRandom	(	double	prob,
		bool	zeroDiagonal = `false`,
		bool	symmetric = `false`
	)

Connect neurons of Clusters with a random connections with the passed probability.

And disconnect neurons with probability (1.0-prob).

Parameters:

prob	is the probability to have a connection between two neurons
zeroDiagonal	if it is true the matrix diagonal is always disconnected
symmetric	if it is true the connections are symmetric