Main Page | Namespace List | Class Hierarchy | Alphabetical List | Class List | File List | Namespace Members | Class Members | File Members | Related Pages

CenterNeuron Class Reference

A center-neuron, the building block of a Radial basis function network. More...

#include <CenterNeuron.h>

Inheritance diagram for CenterNeuron:

Inheritance graph
[legend]
Collaboration diagram for CenterNeuron:

Collaboration graph
[legend]
List of all members.

Public Member Functions

 CenterNeuron (int label, int dimension)
 Constructs a center with the given label and with the center point a random point in the given dimensional space.

 CenterNeuron (int label, Vector center)
 Constructs a center with the given label and given center.

 CenterNeuron (int label, int dimension, real center[])
 Constructs a center with the given label and given center.

virtual ~CenterNeuron ()
Vector getCenter () const
 Returns the center point.

virtual void setCenter (Vector center)
 Sets the center point.

virtual void setCenter (real center[])
 Sets the center point.

virtual void connect (Neuron *from)
 Sets the center-neuron to receive as input the output of the given neuron.

virtual operator std::string () const
 Neuron::string().

virtual const char * getClassName () const
 Returns "CenterNeuron".

virtual void setActivationFunction (ActivationFunction f, ActivationFunction df)
 Sets the activation function of the neuron and its derivative.

virtual uint getDimension () const
 The dimension of the center.


Protected Member Functions

virtual void _recacheOutput () const
 Neuron::_recacheOutput.

virtual void _recacheError () const
 Neuron::_recacheError.


Protected Attributes

real_center
 The center point.

int _dimension
 The dimension of the center point = the size of the input vector.

ActivationFunction _dActivationFunction
 Derivative of the activation function.


Detailed Description

A center-neuron, the building block of a Radial basis function network.

Neurons of this type have a "center" which is a D-dimensional point in space, where D = the number of inputs taken by the neuron. The activation of this neuron is the euclidean distance between the input vector and the center point. The output of this neuron is typically the result of the gaussian distribution function applied to the activation

Todo:
Implement gradient-descent rule based updation of center-point


Constructor & Destructor Documentation

CenterNeuron int  label,
int  dimension
 

Constructs a center with the given label and with the center point a random point in the given dimensional space.

Parameters:
label The label to be given to the neuron
dimension The dimension of the center-point = size of input vector

CenterNeuron int  label,
Vector  center
 

Constructs a center with the given label and given center.

Parameters:
label The label to be given to the neuron
center The center point

CenterNeuron int  label,
int  dimension,
real  center[]
 

Constructs a center with the given label and given center.

Parameters:
label The label to be given to the neuron
dimension The dimension of the center-point = size of input vector
center The center point

virtual ~CenterNeuron  )  [virtual]
 


Member Function Documentation

virtual void _recacheError  )  const [protected, virtual]
 

Neuron::_recacheError.

Implements Neuron.

virtual void _recacheOutput  )  const [protected, virtual]
 

Neuron::_recacheOutput.

Implements Neuron.

virtual void connect Neuron from  )  [virtual]
 

Sets the center-neuron to receive as input the output of the given neuron.

Vector getCenter  )  const
 

Returns the center point.

virtual const char* getClassName  )  const [virtual]
 

Returns "CenterNeuron".

Implements Neuron.

virtual uint getDimension  )  const [virtual]
 

The dimension of the center.

virtual operator std::string  )  const [virtual]
 

Neuron::string().

Reimplemented from Neuron.

virtual void setActivationFunction ActivationFunction  f,
ActivationFunction  df
[virtual]
 

Sets the activation function of the neuron and its derivative.

Parameters:
f The activation function to be used (gaussian by default)
df The derivative of the activation function (dgaussian by default). Required for gradient descent training which is NOT YET IMPLEMENTED

virtual void setCenter real  center[]  )  [virtual]
 

Sets the center point.

virtual void setCenter Vector  center  )  [virtual]
 

Sets the center point.


Member Data Documentation

real* _center [protected]
 

The center point.

ActivationFunction _dActivationFunction [protected]
 

Derivative of the activation function.

Used for gradient descent rule based updation of weights and centers. Not yet implemented

int _dimension [protected]
 

The dimension of the center point = the size of the input vector.


The documentation for this class was generated from the following file:
Generated on Fri Jun 18 13:18:53 2004 for Annie by doxygen 1.3.5