The structure of an RBM is defined by the neuron type of its layers and its interaction parameters. More...

#include <shark/Unsupervised/RBM/Impl/RBMStructure.h>

Inheritance diagram for shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >:

Collaboration diagram for shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >:

Public Types
typedef VectorT	VectorType

typedef VectorMatrixTraits < VectorType >::MatrixType	MatrixType

Public Member Functions
void	read (InArchive &archive)
	Reads the network from an archive. More...

void	write (OutArchive &archive) const
	Writes the network to an archive. More...

void	configure (const PropertyTree &node)
	Configures the network. More...

size_t	numberOfParameters () const
	Returns the total number of parameters of the model. More...

RealVector	parameterVector () const
	Returns the parameters of the RBM as ParameterVector. More...

void	setParameterVector (const RealVector &newParameters)
	Sets the parameters of the RBM. More...

void	setStructure (size_t numOfVN, size_t numOfHN)
	Creates an RBM with a specified number of visible and hidden units. More...

const RealMatrix &	weightMatrix (std::size_t i, std::size_t j) const
	Returns the weight matrix corresponding to the i-th phi-functions of the hidden an the j-th phi-function of the visible variables. More...

RealMatrix &	weightMatrix (std::size_t i, std::size_t j)
	Returns the weight matrix corresponding to the i-th phi-functions of the hidden an the j-th phi-function of the visible variables. More...

std::size_t	numberOfHiddenFeatures ()
	Returns the number of phi-functions of the hidden neuron layer. More...

std::size_t	numberOfVisibleFeatures ()
	Returns the number of phi-functions of the hidden neuron layer. More...

HiddenType &	hiddenNeurons ()
	Returns the layer of hidden neurons. More...

const HiddenType &	hiddenNeurons () const
	Returns the layer of hidden neurons. More...

VisibleType &	visibleNeurons ()
	Returns the layer of visible neurons. More...

const VisibleType &	visibleNeurons () const
	Returns the layer of visible neurons. More...

size_t	numberOfHN () const
	Returns the number of hidden neurons. More...

size_t	numberOfVN () const
	Returns the number of visible neurons. More...

Public Member Functions inherited from shark::IParameterizable
virtual	~IParameterizable ()

Public Member Functions inherited from shark::ISerializable
virtual	~ISerializable ()
	Virtual d'tor. More...

void	load (InArchive &archive, unsigned int version)
	Versioned loading of components, calls read(...). More...

void	save (OutArchive &archive, unsigned int version) const
	Versioned storing of components, calls write(...). More...

	BOOST_SERIALIZATION_SPLIT_MEMBER ()

Public Member Functions inherited from shark::IConfigurable
virtual	~IConfigurable ()
	Virtual d'tor. More...

Protected Attributes
boost::array< MatrixType, hiddenFeatures *visibleFeatures >	m_weightMatrix

HiddenType	m_hiddenNeurons
	The layer of hidden Neurons. More...

VisibleType	m_visibleNeurons
	The Layer of visible Neurons. More...

Detailed Description

template<class HiddenType, class VisibleType, class VectorT, std::size_t hiddenFeatures, std::size_t visibleFeatures>
class shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >

The structure of an RBM is defined by the neuron type of its layers and its interaction parameters.

Definition at line 41 of file RBMStructure.h.

Member Typedef Documentation

template<class HiddenType , class VisibleType , class VectorT , std::size_t hiddenFeatures, std::size_t visibleFeatures>

typedef VectorMatrixTraits<VectorType>::MatrixType shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::MatrixType

Definition at line 44 of file RBMStructure.h.

template<class HiddenType , class VisibleType , class VectorT , std::size_t hiddenFeatures, std::size_t visibleFeatures>

typedef VectorT shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::VectorType

Definition at line 43 of file RBMStructure.h.

Member Function Documentation

template<class HiddenType , class VisibleType , class VectorT , std::size_t hiddenFeatures, std::size_t visibleFeatures>

void shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::configure ( const PropertyTree & node )

inlinevirtual

Configures the network.

two properties must be avaiable: numberOfHN :number of hidden neurons numberOfVN :number of visible neurons

Reimplemented from shark::IConfigurable.

Definition at line 80 of file RBMStructure.h.

References shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::numberOfHN(), shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::numberOfVN(), and shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::setStructure().

template<class HiddenType , class VisibleType , class VectorT , std::size_t hiddenFeatures, std::size_t visibleFeatures>

HiddenType& shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::hiddenNeurons ( )

inline

Returns the layer of hidden neurons.

Definition at line 151 of file RBMStructure.h.

References shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::m_hiddenNeurons.

template<class HiddenType , class VisibleType , class VectorT , std::size_t hiddenFeatures, std::size_t visibleFeatures>

const HiddenType& shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::hiddenNeurons ( ) const

inline

Returns the layer of hidden neurons.

Definition at line 156 of file RBMStructure.h.

References shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::m_hiddenNeurons.

template<class HiddenType , class VisibleType , class VectorT , std::size_t hiddenFeatures, std::size_t visibleFeatures>

std::size_t shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::numberOfHiddenFeatures ( )

inline

Returns the number of phi-functions of the hidden neuron layer.

Definition at line 141 of file RBMStructure.h.

template<class HiddenType , class VisibleType , class VectorT , std::size_t hiddenFeatures, std::size_t visibleFeatures>

size_t shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::numberOfHN ( ) const

inline

Returns the number of hidden neurons.

Definition at line 171 of file RBMStructure.h.

References shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::m_hiddenNeurons.

Referenced by shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::configure(), and shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::numberOfParameters().

template<class HiddenType , class VisibleType , class VectorT , std::size_t hiddenFeatures, std::size_t visibleFeatures>

size_t shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::numberOfParameters ( ) const

inlinevirtual

Returns the total number of parameters of the model.

Reimplemented from shark::IParameterizable.

Definition at line 88 of file RBMStructure.h.

References shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::m_hiddenNeurons, shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::m_visibleNeurons, shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::numberOfHN(), shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::numberOfVN(), and shark::parameters().

Referenced by shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::parameterVector().

template<class HiddenType , class VisibleType , class VectorT , std::size_t hiddenFeatures, std::size_t visibleFeatures>

std::size_t shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::numberOfVisibleFeatures ( )

inline

Returns the number of phi-functions of the hidden neuron layer.

Definition at line 146 of file RBMStructure.h.

template<class HiddenType , class VisibleType , class VectorT , std::size_t hiddenFeatures, std::size_t visibleFeatures>

size_t shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::numberOfVN ( ) const

inline

Returns the number of visible neurons.

Definition at line 176 of file RBMStructure.h.

References shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::m_visibleNeurons.

Referenced by shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::configure(), and shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::numberOfParameters().

template<class HiddenType , class VisibleType , class VectorT , std::size_t hiddenFeatures, std::size_t visibleFeatures>

RealVector shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::parameterVector ( ) const

inlinevirtual

Returns the parameters of the RBM as ParameterVector.

Reimplemented from shark::IParameterizable.

Definition at line 96 of file RBMStructure.h.

References shark::init(), shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::m_hiddenNeurons, shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::m_visibleNeurons, shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::m_weightMatrix, shark::matrixSet(), shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::numberOfParameters(), and shark::parameters().

template<class HiddenType , class VisibleType , class VectorT , std::size_t hiddenFeatures, std::size_t visibleFeatures>

void shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::read ( InArchive & archive )

inlinevirtual

Reads the network from an archive.

Reimplemented from shark::ISerializable.

Definition at line 60 of file RBMStructure.h.

References shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::m_hiddenNeurons, shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::m_visibleNeurons, and shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::m_weightMatrix.

template<class HiddenType , class VisibleType , class VectorT , std::size_t hiddenFeatures, std::size_t visibleFeatures>

void shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::setParameterVector ( const RealVector & newParameters )

inlinevirtual

Sets the parameters of the RBM.

Reimplemented from shark::IParameterizable.

Definition at line 104 of file RBMStructure.h.

References shark::init(), shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::m_hiddenNeurons, shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::m_visibleNeurons, shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::m_weightMatrix, shark::matrixSet(), and shark::parameters().

template<class HiddenType , class VisibleType , class VectorT , std::size_t hiddenFeatures, std::size_t visibleFeatures>

void shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::setStructure	(	size_t	numOfVN,
		size_t	numOfHN
	)

inline

Creates an RBM with a specified number of visible and hidden units.

Parameters

numOfVN	number of visible units
numOfHN	number of hidden units

Definition at line 112 of file RBMStructure.h.

References shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::m_hiddenNeurons, shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::m_visibleNeurons, and shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::m_weightMatrix.

Referenced by shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::configure().

template<class HiddenType , class VisibleType , class VectorT , std::size_t hiddenFeatures, std::size_t visibleFeatures>

VisibleType& shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::visibleNeurons ( )

inline

Returns the layer of visible neurons.

Definition at line 161 of file RBMStructure.h.

References shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::m_visibleNeurons.

template<class HiddenType , class VisibleType , class VectorT , std::size_t hiddenFeatures, std::size_t visibleFeatures>

const VisibleType& shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::visibleNeurons ( ) const

inline

Returns the layer of visible neurons.

Definition at line 166 of file RBMStructure.h.

References shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::m_visibleNeurons.

template<class HiddenType , class VisibleType , class VectorT , std::size_t hiddenFeatures, std::size_t visibleFeatures>

const RealMatrix& shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::weightMatrix	(	std::size_t	i,
		std::size_t	j
	)		const

inline

Returns the weight matrix corresponding to the i-th phi-functions of the hidden an the j-th phi-function of the visible variables.

Parameters

i	index of the hidden phi-function
j	index of the visible phi-function

Definition at line 127 of file RBMStructure.h.

References shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::m_weightMatrix.

template<class HiddenType , class VisibleType , class VectorT , std::size_t hiddenFeatures, std::size_t visibleFeatures>

RealMatrix& shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::weightMatrix	(	std::size_t	i,
		std::size_t	j
	)

inline

Returns the weight matrix corresponding to the i-th phi-functions of the hidden an the j-th phi-function of the visible variables.

Parameters

i	index of the hidden phi-function
j	index of the visible phi-function

Definition at line 136 of file RBMStructure.h.

References shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::m_weightMatrix.

template<class HiddenType , class VisibleType , class VectorT , std::size_t hiddenFeatures, std::size_t visibleFeatures>

void shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::write ( OutArchive & archive ) const

inlinevirtual

Writes the network to an archive.

Reimplemented from shark::ISerializable.

Definition at line 68 of file RBMStructure.h.

References shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::m_hiddenNeurons, shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::m_visibleNeurons, and shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::m_weightMatrix.

Member Data Documentation

template<class HiddenType , class VisibleType , class VectorT , std::size_t hiddenFeatures, std::size_t visibleFeatures>

HiddenType shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::m_hiddenNeurons

protected

The layer of hidden Neurons.

Definition at line 52 of file RBMStructure.h.

template<class HiddenType , class VisibleType , class VectorT , std::size_t hiddenFeatures, std::size_t visibleFeatures>

VisibleType shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::m_visibleNeurons

protected

The Layer of visible Neurons.

Definition at line 55 of file RBMStructure.h.

template<class HiddenType , class VisibleType , class VectorT , std::size_t hiddenFeatures, std::size_t visibleFeatures>

boost::array<MatrixType,hiddenFeatures * visibleFeatures> shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::m_weightMatrix

protected

Matrices that identify the weights associated with the connections between the visible and the hidden neurons (In the standard case just one single matrix)

Definition at line 49 of file RBMStructure.h.

Referenced by shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::parameterVector(), shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::read(), shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::setParameterVector(), shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::setStructure(), shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::weightMatrix(), and shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >::write().

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

include/shark/Unsupervised/RBM/Impl/RBMStructure.h

Public Types

Public Member Functions

Protected Attributes

Detailed Description

template<class HiddenType, class VisibleType, class VectorT, std::size_t hiddenFeatures, std::size_t visibleFeatures> class shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >

Member Typedef Documentation

Member Function Documentation

Member Data Documentation

template<class HiddenType, class VisibleType, class VectorT, std::size_t hiddenFeatures, std::size_t visibleFeatures>
class shark::RBMStructure< HiddenType, VisibleType, VectorT, hiddenFeatures, visibleFeatures >