KohonenNetwork Class Reference

Todo: getOutput (&& getWin*) doesn't give anything sensible in this impl. More...

#include <Kohonen.h>

Inheritance diagram for KohonenNetwork:

[legend]Collaboration diagram for KohonenNetwork:

[legend]List of all members.


Public Types
enum	{ INPUT_LAYER, OUTPUT_LAYER }
Public Member Functions
	KohonenNetwork (uint inputs, const Topology &topology, KohonenParameters &params, PublicValues &ctrl=defaultControl)
	default LP threshold
void	train (TrainingSet &T, real terminatingLP=TERMINATING_LP)
	Train the net until learningParameter reaches given threshold.
void	trainExample (const Vector &input)
	alternative to training via train. You can train one-by-one example directly, but should must call iterateTime() to control the dynamic
uint	getWinnerOutput (const Vector &input)
	version of getOutput for Kohonen
uint	getWinner () const
	version of getWinnerOutput based on the current state of network.
virtual Vector	getOutput (const Vector &input)
	Warning - this raw output function doesn't give much sense!
const Layer &	getOutputLayer () const
virtual const char *	getClassName () const
	For reflection.
virtual Vector	getCluster (uint cluserNum) const
	Get vector representing the given cluster - input weights of given neuron.
virtual void	save (const std::string &filename)
	Save the network structure to the given text file.
const Topology &	getTopology () const
Static Public Attributes
const real	MAX_OUTPUT
const real	MIN_OUTPUT
const real	TERMINATING_LP
Protected Member Functions
void	connectLayers ()
void	randomizeWeights ()
real	getNBSize () const
real	getLearningParam () const
void	_setInput (const Vector &input)
Protected Attributes
InputLayer	_inputLayer
Layer	_outputLayer
const Topology &	topology
KohonenParameters &	parameters
PublicValues &	control
uint	_winner
bool	_winnerValid

Detailed Description

Todo: getOutput (&& getWin*) doesn't give anything sensible in this impl.

!!!!!!!!

Member Enumeration Documentation

anonymous enum

Enumeration values:

INPUT_LAYER

OUTPUT_LAYER

Constructor & Destructor Documentation

KohonenNetwork ( uint inputs,

const Topology & topology,

KohonenParameters & params,

PublicValues & ctrl = defaultControl

)

default LP threshold

Member Function Documentation

void _setInput ( const Vector & input ) [protected]

void connectLayers ( ) [protected]

virtual const char* getClassName ( ) const [inline, virtual]

For reflection.
All sub-classes MUST implement this method, which should just return class name
Implements Network.

virtual Vector getCluster ( uint cluserNum ) const [virtual]

Get vector representing the given cluster - input weights of given neuron.

real getLearningParam ( ) const [inline, protected]

Here is the call graph for this function:

real getNBSize ( ) const [inline, protected]

Here is the call graph for this function:

virtual Vector getOutput ( const Vector & input ) [virtual]

Warning - this raw output function doesn't give much sense!

Implements Network.

const Layer& getOutputLayer ( ) const [inline]

const Topology& getTopology ( ) const [inline]

uint getWinner ( ) const

version of getWinnerOutput based on the current state of network.
Reasonable to call only after previous get*Output

uint getWinnerOutput ( const Vector & input )

version of getOutput for Kohonen

Returns:
the winning neuron's index

void randomizeWeights ( ) [protected]

virtual void save ( const std::string & filename ) [inline, virtual]

Save the network structure to the given text file.
The file format is quite simple and is commented.
Parameters:

filename Name of the file to save network information into

Implements Network.

void train ( TrainingSet & T,

real terminatingLP = TERMINATING_LP

)

Train the net until learningParameter reaches given threshold.

void trainExample ( const Vector & input )

alternative to training via train. You can train one-by-one example directly, but should must call iterateTime() to control the dynamic