lti::stdLayer Class Reference

This abstract class implements a standard neural network layer. More...

#include <ltiStdLayer.h>

Inheritance diagram for lti::stdLayer:

[legend]

Collaboration diagram for lti::stdLayer:

[legend]

List of all members.

Classes
class	activationFunctor
	Parent class for all activation function functors. More...
class	distancePropFunctor
	Distance propagation functor. More...
class	dotPropFunctor
	Dot-Product propagation functor. More...
class	gaussActFunctor
	a gaussian activation function More...
class	initFunctor
	This class and all it's derived classes should initialize the network layer. More...
class	linearActFunctor
	a linear activation function More...
class	propagationFunctor
	Parent class for all propagation functors All propagation functors use a given activation function to calculate the output layer vector using a given input vector and a given weights matrix. More...
class	sigmoidActFunctor
	a sigmoid activation function More...
class	trainFunctor
	training functor More...
Public Member Functions
	stdLayer (const std::string &theName)
	stdLayer (const stdLayer &other)
virtual	~stdLayer ()
stdLayer &	copy (const stdLayer &other)
virtual void	initWeights (initFunctor &initFunc)
virtual bool	train (const dvector &input, const int &trainID, trainFunctor &trainFunc)
virtual bool	train (const dvector &input, trainFunctor &trainFunc)
virtual bool	propagate (const dvector &input, propagationFunctor &prop, dvector &output) const
const ivector &	getOutID () const
void	setSize (const int &in, const int &out)
virtual bool	write (ioHandler &handler, const bool complete=true) const
virtual bool	read (ioHandler &handler, const bool complete=true)
virtual const char *	getStatusString () const
virtual void	setStatusString (const char *msg) const
const dmatrix &	getWeights ()
void	setLearnRate (const double &theLR)


void	setWeights (const dmatrix &weights)
void	setIDs (const ivector &ids)
deprecated methods
Deprecated methods
virtual bool	save (std::ostream &out) const
virtual bool	load (std::istream &in)
virtual bool	saveBinary (std::ostream &out) const
virtual bool	loadBinary (std::istream &in)
Protected Attributes
dmatrix	weights
ivector	outID
int	sizeIn
int	sizeOut
std::string	layerName
double	learnRate
char *	statusString
Static Protected Attributes
static const char *const	emptyString

Detailed Description

This abstract class implements a standard neural network layer.

This class contains the weights matrix, the vector with the output values, the class id for each output element, the dimensionality of the input vector and output layer (which must be consistent with the dimensions of the weights matrix and the output vector), the name of the layer (an ASCII string) and a learn rate value.

The enclosed classes are supponsed to modify this values. There are some initialization, activation, propagation and training functors.

Constructor & Destructor Documentation

lti::stdLayer::stdLayer ( const std::string & theName )

default constructor

lti::stdLayer::stdLayer ( const stdLayer & other )

copy constructor

virtual lti::stdLayer::~stdLayer ( ) [virtual]

destructor

Member Function Documentation

stdLayer& lti::stdLayer::copy ( const stdLayer & other )

copy member

const ivector & lti::stdLayer::getOutID ( ) const [inline]

get the output vector

get the class IDs for output vector elements

References outID.

virtual const char* lti::stdLayer::getStatusString ( ) const [virtual]

return the last message set with setStatusString().

This will never return 0. If no status-string has been set yet an empty string (pointer to a string with only the char(0)) will be returned.

const dmatrix & lti::stdLayer::getWeights ( ) [inline]

get the weights of the layer

References weights.

virtual void lti::stdLayer::initWeights ( initFunctor & initFunc ) [virtual]

initialize the weights

virtual bool lti::stdLayer::load ( std::istream & in ) [virtual]

read the standard layer from the given stream (ascii)

Deprecated:: Use read() instead

virtual bool lti::stdLayer::loadBinary ( std::istream & in ) [virtual]

read the standard layer from the given stream (binary)

Deprecated:: Use read() instead

virtual bool lti::stdLayer::propagate	(	const dvector &	input,
		propagationFunctor &	prop,
		dvector &	output
	)			const `[virtual]`

propagate the vector through the network

Returns:: false if an error occurred, otherwise true

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

read the standard layer 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

virtual bool lti::stdLayer::save ( std::ostream & out ) const [virtual]

write the standard layer in the given stream (ascii)

Deprecated:: Use write() instead

virtual bool lti::stdLayer::saveBinary ( std::ostream & out ) const [virtual]

write the standard layer in the given stream (binary)

Deprecated:: Use write() instead

void lti::stdLayer::setIDs ( const ivector & ids )

Set the IDs of the layer to the given vector.

void lti::stdLayer::setLearnRate ( const double & theLR ) [inline]

set the learn rate

References learnRate.

void lti::stdLayer::setSize	(	const int &	in,
		const int &	out
	)

set the size of the input and output vectors

Parameters:

	in	size of the input vector
	out	size of the output vector

virtual void lti::stdLayer::setStatusString ( const char * msg ) const [virtual]

set a status string.

Parameters:

msg

the const string to be reported next time by getStatusString(). The given string will be copied This message will be usually set within the apply methods to indicate an error cause.

Note that the change of the status string is not considered as a change in the functor status.

void lti::stdLayer::setWeights ( const dmatrix & weights )

These methods should only be used by the Axiom Java interface.

Set the weights of the layer to the given matrix

virtual bool lti::stdLayer::train	(	const dvector &	input,
		trainFunctor &	trainFunc
	)			`[virtual]`

train the network layer (unsupervised) with the given input vector

Returns:: false if an error occurred, otherwise true

virtual bool lti::stdLayer::train	(	const dvector &	input,
		const int &	trainID,
		trainFunctor &	trainFunc
	)			`[virtual]`

train the network layer (supervised) with the given input vector

Does one step of training, i.e. one feature vector 'input' of class 'trainID' is learned by the layer as specified in the trainFunctor

Returns:: false if an error occurred, otherwise true

Reimplemented in lti::rbf::layer1, and lti::rbf::layer2.

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

write the standard layer 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.