RBFNet Class Reference

Offers the functions to create and to work with radial basis function network. More...

#include <RBFNet.h>

Inheritance diagram for RBFNet:

Public Member Functions
void	read (std::istream &is)
	Reads a network from the input stream "is".
void	write (std::ostream &os) const
	Writes a network to the output stream "os".
	RBFNet (unsigned numInput, unsigned numOutput, unsigned numHidden)
	Creates a Radial Basis Function Network.
	RBFNet (const std::string &filename)
	Creates a Radial Basis Function Network by reading information from file "filename".
	RBFNet (unsigned numInput, unsigned numOutput, unsigned numHidden, const Array< double > &_m, const Array< double > &_A, const Array< double > &_b, const Array< double > &_v)
void	initRBFNet (const Array< double > &input, const Array< double > &target)
	Initializes the Radial Basis Function Network.
void	model (const Array< double > &input, Array< double > &target)
	Evaluates the RBFN-Model and calculates the output values.
void	setParameter (unsigned int index, double value)
	Modifies a specific model parameter.
void	modelDerivative (const Array< double > &input, Array< double > &derivative)
	Calculates the derivatives of the model with respect to the weights and stores the results in ModelInterface::dmdw.
void	modelDerivative (const Array< double > &input, Array< double > &output, Array< double > &derivative)
	Calculates the derivatives of the model with respect to the weights and stores the results in ModelInterface::dmdw. Furthermode, the output of the model is evaluated.
const Array< double > &	getWeights () const
	Returns the weights of the connections between hidden and output neurons.
const Array< double > &	getBias () const
	Returns the bias of all output neurons.
const Array< double > &	getCenter () const
	Returns the centers of all hidden neurons.
const Array< double > &	getVariance () const
	Returns the variances of all hidden neurons.
const unsigned	getNHidden () const
	Returns the number of hidden neurons.
void	setWeights (const Array< double > &_A)
	Sets the weights of connections between hidden and output neurons.
void	setBias (const Array< double > &_b)
	Sets the bias of the output neurons.
void	setCenter (const Array< double > &_m)
	Sets the centers of the hidden neurons.
void	setVariance (const Array< double > &_v)
	Sets the variances of the hidden neurons.

Detailed Description

Offers the functions to create and to work with radial basis function network.

Together with an error measure and an optimization algorithm class you can use this radial basis function network class to produce your own optimization tool. Radial Basis Function networks are especially used for classification tasks. For this purpose the class of models contains, besides the number of hidden neurons $nHidden$ , 4 different kinds of parameters: the centers of the hidden neurons $\vec{m}_j=(m_{j1},\ldots,m_{jn})$ , the standard deviations of the hidden neurons $\vec{\sigma}_j=(\sigma_{j1},\ldots,\sigma_{jn})$ , the weights of the connections between the hidden neuron with index $j$ and output neuron with index $k$ and the bias $b_k$ of the output neuron with index $k$ . The output neuron with index $k$ yields the following output, assuming a $nInput$ -dimensional input vector $\vec{x}$ :

$y_k = b_k + \sum_{j=1}^{nHidden} \exp\left[ -\frac{1}{2}\sum_{i=1}^{nInput} \frac{\left(x_i-m_{ji}\right)2}{\sigma_{ji}}\right]$

Author:: C. Igel

Date:: 1999

Changes:: none

Status:: stable

Definition at line 96 of file RBFNet.h.

Constructor & Destructor Documentation

RBFNet::RBFNet	(	unsigned	numInput,
		unsigned	numOutput,
		unsigned	numHidden
	)			`[inline]`

Creates a Radial Basis Function Network.

Constructor 1

This method creates a Radial Basis Function Network (RBFN) with numInput input neurons, numOutput output neurons and numHidden hidden neurons.

Parameters:

	numInput	Number of input neurons, equal to dimensionality of input space.
	numOutput	Number of output neurons, equal to dimensionality of output space.
	numHidden	Number of hidden neurons.

Returns:: None.

Author:: C. Igel

Date:: 1999

Changes: none

Status: none

Definition at line 130 of file RBFNet.h.

References Model::inputDimension, Model::outputDimension, and Model::parameter.

RBFNet::RBFNet ( const std::string & filename ) [inline]

Creates a Radial Basis Function Network by reading information from file "filename".

Constructor 2

This method creates a Radial Basis Function Network (RBFN) which is definded in the file filename.

Parameters:

filename

Name of the file the RBFN definition is read from.

Returns:: None.

Author:: L. Arnold

Date:: 2002

Changes: none

Status: stable

Definition at line 165 of file RBFNet.h.

RBFNet::RBFNet	(	unsigned	numInput,
		unsigned	numOutput,
		unsigned	numHidden,
		const Array< double > &	_m,
		const Array< double > &	_A,
		const Array< double > &	_b,
		const Array< double > &	_v
	)			`[inline]`

Constructor 3

This method creates a Radial Basis Function Network (RBFN) with numInput input neurons, numOutput output neurons and numHidden hidden neurons and initializes the parameters.

Parameters:

	numInput	Number of input neurons, equal to dimensionality of input space.
	numOutput	Number of output neurons, equal to dimensionality of output space.
	numHidden	Number of hidden neurons.
	_m	centers
	_A	output weights
	_b	biases
	_v	variances

Returns:: None.

Author:: C. Igel

Date:: 2003

Changes: none

Status: stable

Definition at line 215 of file RBFNet.h.

References Model::inputDimension, Model::outputDimension, and Model::parameter.

Member Function Documentation

const Array<double>& RBFNet::getBias ( ) const [inline]

Returns the bias of all output neurons.

The bias values are contained in an 1-dimensional array. The element $b_{k}$ is the bias of the $k^{th}$ output neuron.

Returns:: The bias values of the output neurons.

Author:: M. Hüsken

Date:: 2001

Changes: none

Status: stable

Definition at line 404 of file RBFNet.h.

Referenced by write().

const Array<double>& RBFNet::getCenter ( ) const [inline]

Returns the centers of all hidden neurons.

The center values are contained in an 2-dimensional array. The element $m_{ji}$ is the $i^{th}$ coordinate of the center of the $j^{th}$ hidden neuron.

Returns:: The center values of the hidden neurons.

Author:: M. Hüsken

Date:: 2001

Changes: none

Status: stable

Definition at line 430 of file RBFNet.h.

Referenced by write().

const unsigned RBFNet::getNHidden ( ) const [inline]

Returns the number of hidden neurons.

Returns:: The number of hidden neurons.

Author:: M. Hüsken

Date:: 2001

Changes: none

Status: stable

Definition at line 478 of file RBFNet.h.

Referenced by write().

const Array<double>& RBFNet::getVariance ( ) const [inline]

Returns the variances of all hidden neurons.

The variances are contained in an 2-dimensional array. The element $\sigma_{ji}$ is the $i^{th}$ coordinate of the variance of the $j^{th}$ hidden neuron.

Returns:: The variances of the hidden neurons.

Author:: M. Hüsken

Date:: 2001

Changes: none

Status: stable

Definition at line 456 of file RBFNet.h.

Referenced by write().

const Array<double>& RBFNet::getWeights ( ) const [inline]

Returns the weights of the connections between hidden and output neurons.

Returns an array containing the weights of the connections between hidden and output neurons. The array is a two-dimensional object, the element $w_{ji}$ is the weight of the connection from the $i^{th}$ hidden neuron to the $j^{th}$ output neuron.

Returns:: The weight matrix.

Author:: M. Hüsken

Date:: 2001

Changes: none

Status: stable

Definition at line 378 of file RBFNet.h.

Referenced by write().

void RBFNet::initRBFNet	(	const Array< double > &	input,
		const Array< double > &	target
	)			`[inline]`

Initializes the Radial Basis Function Network.

This method initializes the parameters of the RBFN-model. First, it performs a K-means-Clustering to determine the centers and variances of the hidden neurons. Thereafter, the connection weights and bias values are determined by means of a linear regression.

Parameters:

	input	Vector of input values.
	target	Vector of target values.

Returns:: None.

Author:: C. Igel

Date:: 1999

Changes: none

Status: none

Definition at line 260 of file RBFNet.h.

References Model::parameter.

void RBFNet::model	(	const Array< double > &	input,
		Array< double > &	target
	)			`[inline, virtual]`

Evaluates the RBFN-Model and calculates the output values.

This method evaluates the RBFN and determines the output values based on the following formula:

$y_k = b_k + \sum_{j=1}^{nHidden} \exp\left[ -\frac{1}{2}\sum_{i=1}^{nInput} \frac{\left(x_i-m_{ji}\right)2}{\sigma_{ji}}\right]$

Parameters:

	input	Vector of input values.
	target	Vector of target values.

Returns:: None.

Author:: C. Igel

Date:: 1999

Changes: none

Status: none

Implements Model.

Definition at line 292 of file RBFNet.h.

References Model::parameter.

void RBFNet::modelDerivative	(	const Array< double > &	input,
		Array< double > &	output,
		Array< double > &	derivative
	)			`[inline, virtual]`

Calculates the derivatives of the model with respect to the weights and stores the results in ModelInterface::dmdw. Furthermode, the output of the model is evaluated.

Parameters:

	input	Vector of input values.
	output	Vector of evaluated output values.

Returns:: None.

Author:: C. Igel

Date:: 1999

Changes: none

Status: none

Reimplemented from Model.

Definition at line 349 of file RBFNet.h.

References Model::parameter.

void RBFNet::modelDerivative	(	const Array< double > &	input,
		Array< double > &	derivative
	)			`[inline, virtual]`

Calculates the derivatives of the model with respect to the weights and stores the results in ModelInterface::dmdw.

Parameters:

input

Vector of input values.

Returns:: None.

Author:: C. Igel

Date:: 1999

Changes: none

Status: none

Reimplemented from Model.

Definition at line 323 of file RBFNet.h.

References Model::parameter.

void RBFNet::read ( std::istream & is ) [virtual]

Reads a network from the input stream "is".

The number of input, output and hidden neurons, and also the center, variance, weight and bias matrices are read from the input stream is, memory for the internal net data structures is reserved and the weight and bias values are written to the parameter vector Model::parameter.

Parameters:

is

The input stream.

Returns:: Reference to the input stream.

Author:: L. Arnold

Date:: 2002

Changes: none

Status: stable

Reimplemented from Model.

Definition at line 668 of file RBFNet.h.

References i, Model::inputDimension, Model::outputDimension, Model::parameter, setBias(), setCenter(), setVariance(), and setWeights().

void RBFNet::setBias ( const Array< double > & _b ) [inline]

Sets the bias of the output neurons.

This method sets the bias of the output neurons. The bias is described by means of an one-dimensional array, the element $b_{k}$ is the bias of the $k^{th}$ output neuron.

Parameters:

_b

The new bias values.

Returns:: None.

Author:: M. Hüsken

Date:: 2001

Changes: none

Status: stable

Definition at line 535 of file RBFNet.h.

References Model::parameter.

Referenced by read().

void RBFNet::setCenter ( const Array< double > & _m ) [inline]

Sets the centers of the hidden neurons.

This method sets the centers of the hidden neurons. The centers are described by means of a two-dimensional array, the element $m_{ji}$ is the $i^{th}$ coordinate of the center of the $j^{th}$ hidden neuron.

Parameters:

_m

The new center values.

Returns:: None.

Author:: M. Hüsken

Date:: 2001

Changes: none

Status: stable

Definition at line 564 of file RBFNet.h.

References Model::parameter.

Referenced by read().

void RBFNet::setParameter	(	unsigned int	index,
		double	value
	)			`[inline, virtual]`

Modifies a specific model parameter.

Reimplemented from Model.

Definition at line 299 of file RBFNet.h.

References Model::parameter.

void RBFNet::setVariance ( const Array< double > & _v ) [inline]

Sets the variances of the hidden neurons.

This method sets the variances of the hidden neurons. The variances are described by means of a two-dimensional array, the element $\sigma_{ji}$ is the $i^{th}$ coordinate of the variance of the $j^{th}$ hidden neuron.

Parameters:

_v

The new variance values.

Returns:: None.

Author:: M. Hüsken

Date:: 2001

Changes: none

Status: stable

Definition at line 593 of file RBFNet.h.

References Model::parameter.

Referenced by read().

void RBFNet::setWeights ( const Array< double > & _A ) [inline]

Sets the weights of connections between hidden and output neurons.

Sets the weights of the connections between hidden and output neurons. The weights are described by means of a two-dimensional array, the element $w_{ji}$ is the weight of the connection from the $i^{th}$ hidden neuron to the $j^{th}$ output neuron.

Parameters:

_A

The new weight values.

Returns:: None.

Author:: M. Hüsken

Date:: 2001

Changes: none

Status: stable

Definition at line 507 of file RBFNet.h.

References Model::parameter.

Referenced by read().

void RBFNet::write ( std::ostream & os ) const [virtual]

Writes a network to the output stream "os".

The number of input, output and hidden neurons and also the center, variance, weights and bias matrices are written to the output stream os.

Parameters:

	os	The output stream.
	net	The net that will be written to the output stream.

Returns:: Reference to the output stream.

Author:: L. Arnold

Date:: 2002

Changes: none

Status: stable

Reimplemented from Model.

Definition at line 628 of file RBFNet.h.

References getBias(), getCenter(), getNHidden(), getVariance(), and getWeights().

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

RBFNet.h

RBFNet Class Reference

Public Member Functions

Detailed Description

Constructor & Destructor Documentation

Member Function Documentation