lti::SOFM::parameters Class Reference

the parameters for the class SOFM More...

#include <ltiSOFM.h>

Inheritance diagram for lti::SOFM::parameters:

[legend]

Collaboration diagram for lti::SOFM::parameters:

[legend]

List of all members.

Public Types
enum	eInit { Random = 1, Linear = 2 }
enum	eMetric { L1 = 1, L2, Dot }
Public Member Functions
	parameters ()
	parameters (const parameters &other)
virtual	~parameters ()
const char *	getTypeName () const
parameters &	copy (const parameters &other)
parameters &	operator= (const parameters &other)
virtual classifier::parameters *	clone () const
virtual bool	write (ioHandler &handler, const bool complete=true) const
virtual bool	read (ioHandler &handler, const bool complete=true)
Public Attributes
eInit	initType
eMetric	metricType
int	stepsOrdering
double	orderLearnMax
double	orderLearnMin
double	orderNeighborMax
double	orderNeighborMin
double	orderNeighborThresh
int	stepsConvergence
double	convergenceLearnMax
double	convergenceLearnMin

---

Detailed Description

the parameters for the class SOFM

---

Member Enumeration Documentation

enum lti::SOFM::parameters::eInit

Possible types of initialization for the grid of neurons:.

Random

Grid points are selected randomly from the data.

Linear

The mean as well as the two maximum eigenvalues and corresponding eigenvectors of the data are calculted. Then the neurons in the grid are uniformly spread over the rectangle spanned by the eigenvectors and of the dimensions of the eigenvalues. Like this the grid is already ordered and neurons should be close to their final destination. Here, it is useful to set the parameter calculateSize to true

Enumerator:

Random	choose initial values randomly from data
Linear	take max eigenvalues and mean and distribute points

enum lti::SOFM::parameters::eMetric

Different metrics can be used to determine the distance between an input vector and each neuron.

Enumerator:

L1	use L1 norm
L2	use L2 norm
Dot	use dot product

---

Constructor & Destructor Documentation

lti::SOFM::parameters::parameters

(

)

default constructor

Reimplemented from lti::unsupervisedClassifier::parameters.

Reimplemented in lti::SOFM2D::parameters.

lti::SOFM::parameters::parameters

(

const parameters &

other

)

copy constructor

Parameters:

other

the parameters object to be copied

Reimplemented from lti::unsupervisedClassifier::parameters.

Reimplemented in lti::SOFM2D::parameters.

virtual lti::SOFM::parameters::~parameters

(

)

[virtual]

destructor

Reimplemented from lti::unsupervisedClassifier::parameters.

Reimplemented in lti::SOFM2D::parameters.

---

Member Function Documentation

virtual classifier::parameters* lti::SOFM::parameters::clone

(

)

const [virtual]

returns a pointer to a clone of the parameters

Reimplemented from lti::classifier::parameters.

Reimplemented in lti::SOFM2D::parameters.

parameters& lti::SOFM::parameters::copy

(

const parameters &

other

)

copy the contents of a parameters object

Parameters:

other

the parameters object to be copied

Returns:

a reference to this parameters object

Reimplemented from lti::unsupervisedClassifier::parameters.

Reimplemented in lti::SOFM2D::parameters.

const char* lti::SOFM::parameters::getTypeName

(

)

const [virtual]

returns name of this type

Reimplemented from lti::unsupervisedClassifier::parameters.

Reimplemented in lti::SOFM2D::parameters.

parameters& lti::SOFM::parameters::operator=

(

const parameters &

other

)

copy the contents of a parameters object

Parameters:

other

the parameters object to be copied

Returns:

a reference to this parameters object

Reimplemented from lti::unsupervisedClassifier::parameters.

Reimplemented in lti::SOFM2D::parameters.

virtual bool lti::SOFM::parameters::read	(	ioHandler &	handler,
		const bool	complete = true
	)			[virtual]

read the parameters from the given ioHandler

Parameters:

	handler	the ioHandler to be used
	complete	if true (the default) the enclosing begin/end will be also written, otherwise only the data block will be written.

Returns:

true if write was successful

Reimplemented from lti::classifier::parameters.

Reimplemented in lti::SOFM2D::parameters.

virtual bool lti::SOFM::parameters::write	(	ioHandler &	handler,
		const bool	complete = true
	)			const [virtual]

write the parameters in the given ioHandler

Parameters:

	handler	the ioHandler to be used
	complete	if true (the default) the enclosing begin/end will be also written, otherwise only the data block will be written.

Returns:

true if write was successful

Reimplemented from lti::classifier::parameters.

Reimplemented in lti::SOFM2D::parameters.

---

Member Data Documentation

double lti::SOFM::parameters::convergenceLearnMax

Maximum value of learn-rate during convergence.

double lti::SOFM::parameters::convergenceLearnMin

Minimum value of learn-rate during convergence.

eInit lti::SOFM::parameters::initType

type of initialization method used.

See eInit.

eMetric lti::SOFM::parameters::metricType

type of metric used.

See eMetric.

double lti::SOFM::parameters::orderLearnMax

Maximum value of learn-rate during ordering.

double lti::SOFM::parameters::orderLearnMin

Minimum value of learn-rate during ordering.

double lti::SOFM::parameters::orderNeighborMax

Maximum distance to the winner neuron during ordering.

The unit is in grid steps. E.g. a value of 1 would include (theoretically) the 8 neighbors, since a square window is assumed. You can change this by setting orderNeighborThresh to an appropriate value.

double lti::SOFM::parameters::orderNeighborMin

Minimum distance to the winner neuron during ordering.

The neighborhood stays like this for the convergence phase. This value should be rather low, typically 1. For values below 1 no neighborhood is created which will not work!

double lti::SOFM::parameters::orderNeighborThresh

The minimum value the neighborhood function assumes (cut-off value).

This should be rather low. Since a Gaussian kernel is used this value steers the size of the kernel to be used. I.e. the bigger the value, the faster the training and the less precise. E.g. if for the current neighborhood size the Gaussian value is smaller than the threshold for the roof of the neighborhood value, the kernel size will be lower than the roof of the neighborhood size.

int lti::SOFM::parameters::stepsConvergence

Number of epochs used for convergence in the training.

int lti::SOFM::parameters::stepsOrdering

Number of epochs used for ordering in the training.

---

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

• ltiSOFM.h