Base class for all Models. More...

#include <shark/Models/AbstractModel.h>

Inheritance diagram for shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >:

Public Types
enum	Feature { HAS_FIRST_PARAMETER_DERIVATIVE = 1 , HAS_FIRST_INPUT_DERIVATIVE = 4 }

typedef InputTypeT	InputType
	Defines the input type of the model.

typedef OutputTypeT	OutputType
	Defines the output type of the model.

typedef AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >	ModelBaseType
	Defines the BaseType used by the model (this type). Useful for creating derived models.

typedef Batch< InputType >::type	BatchInputType
	defines the batch type of the input type.

typedef Batch< OutputType >::type	BatchOutputType
	defines the batch type of the output type

typedef TypedFlags< Feature >	Features

typedef TypedFeatureNotAvailableException< Feature >	FeatureNotAvailableException

Public Types inherited from shark::IParameterizable< VectorType >
typedef VectorType	ParameterVectorType

Public Member Functions
	AbstractModel ()

virtual	~AbstractModel ()

const Features &	features () const

virtual void	updateFeatures ()

bool	hasFirstParameterDerivative () const
	Returns true when the first parameter derivative is implemented.

bool	hasFirstInputDerivative () const
	Returns true when the first input derivative is implemented.

virtual Shape	inputShape () const =0
	Returns the expected shape of the input.

virtual Shape	outputShape () const =0
	Returns the shape of the output.

virtual boost::shared_ptr< State >	createState () const
	Creates an internal state of the model.

virtual void	read (InArchive &archive)
	From ISerializable, reads a model from an archive.

virtual void	write (OutArchive &archive) const
	writes a model to an archive

virtual void	eval (BatchInputType const &patterns, BatchOutputType &outputs) const
	Standard interface for evaluating the response of the model to a batch of patterns.

virtual void	eval (BatchInputType const &patterns, BatchOutputType &outputs, State &state) const =0
	Standard interface for evaluating the response of the model to a batch of patterns.

virtual void	eval (InputType const &pattern, OutputType &output) const
	Standard interface for evaluating the response of the model to a single pattern.

Data< OutputType >	operator() (Data< InputType > const &patterns) const
	Model evaluation as an operator for a whole dataset. This is a convenience function.

OutputType	operator() (InputType const &pattern) const
	Model evaluation as an operator for a single pattern. This is a convenience function.

BatchOutputType	operator() (BatchInputType const &patterns) const
	Model evaluation as an operator for a single pattern. This is a convenience function.

virtual void	weightedParameterDerivative (BatchInputType const &pattern, BatchOutputType const &outputs, BatchOutputType const &coefficients, State const &state, ParameterVectorType &derivative) const
	calculates the weighted sum of derivatives w.r.t the parameters.

virtual void	weightedInputDerivative (BatchInputType const &pattern, BatchOutputType const &outputs, BatchOutputType const &coefficients, State const &state, BatchInputType &derivative) const
	calculates the weighted sum of derivatives w.r.t the inputs

virtual void	weightedDerivatives (BatchInputType const &patterns, BatchOutputType const &outputs, BatchOutputType const &coefficients, State const &state, ParameterVectorType &parameterDerivative, BatchInputType &inputDerivative) const
	calculates weighted input and parameter derivative at the same time

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

virtual ParameterVectorType	parameterVector () const
	Return the parameter vector.

virtual void	setParameterVector (ParameterVectorType const &newParameters)
	Set the parameter vector.

virtual std::size_t	numberOfParameters () const
	Return the number of parameters.

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

virtual std::string	name () const
	returns the name of the object

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

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

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

	BOOST_SERIALIZATION_SPLIT_MEMBER ()

Protected Attributes
Features	m_features

Detailed Description

template<class InputTypeT, class OutputTypeT, class ParameterVectorType = RealVector>
class shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >

Base class for all Models.

: A model is one of the three fundaments of supervised learning: model, error measure and an optimization algorithm. It is a concept of a function which performs a mapping \( x \rightarrow f_w(x)\). In contrast to an error function it has two sets of parameters: The first is the current point to map \(x\), the others are the internal model parameters \(w\) which define the mapping. Often a model is used to find an optimal mapping for a problem, for example a function which best fits the points of a given dataset. Therefore, AbstractModel does not only offer the mapping itself, but also a set of special derivatives with respect to \( x \) and \( w \). Most of the time, only the derivative with respect to \( w \) is needed, but in some special problems, like finding optimal stimuli or stacking models, also the input derivative is needed.

Models are optimized for batch processing. This means, that instead of only one data point at a time, it can: evaluate a big set of inputs at the same time, using optimized routines for this task.

: The derivatives are weighted, which means that the derivatives of every single output are added together weighted by coefficients (see weightedParameterDerivative). This is an optimization for the chain rule which is very efficient to calculate most of the time.

: It is allowed to store intermediate values during eval and use them to speed up calculation of derivatives. Therefore it must be guaranteed that eval() is called before calculating derivatives. This is no restriction, since typical error measures need the mapping itself and not only the derivative.

: Models have names and can be serialised and have parameters. The type of the parameter vector can be set as third argument. By default, this is RealVector.

Definition at line 81 of file AbstractModel.h.

Member Typedef Documentation

◆ BatchInputType

template<class InputTypeT , class OutputTypeT , class ParameterVectorType = RealVector>

typedef Batch<InputType>::type shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::BatchInputType

defines the batch type of the input type.

This could for example be std::vector<InputType> but for example for RealVector it could be RealMatrix

Definition at line 97 of file AbstractModel.h.

◆ BatchOutputType

template<class InputTypeT , class OutputTypeT , class ParameterVectorType = RealVector>

typedef Batch<OutputType>::type shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::BatchOutputType

defines the batch type of the output type

Definition at line 99 of file AbstractModel.h.

◆ FeatureNotAvailableException

template<class InputTypeT , class OutputTypeT , class ParameterVectorType = RealVector>

typedef TypedFeatureNotAvailableException<Feature> shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::FeatureNotAvailableException

Definition at line 110 of file AbstractModel.h.

◆ Features

template<class InputTypeT , class OutputTypeT , class ParameterVectorType = RealVector>

typedef TypedFlags<Feature> shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::Features

Definition at line 110 of file AbstractModel.h.

◆ InputType

template<class InputTypeT , class OutputTypeT , class ParameterVectorType = RealVector>

typedef InputTypeT shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::InputType

Defines the input type of the model.

Definition at line 85 of file AbstractModel.h.

◆ ModelBaseType

template<class InputTypeT , class OutputTypeT , class ParameterVectorType = RealVector>

typedef AbstractModel<InputTypeT,OutputTypeT,ParameterVectorType> shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::ModelBaseType

Defines the BaseType used by the model (this type). Useful for creating derived models.

Definition at line 92 of file AbstractModel.h.

◆ OutputType

template<class InputTypeT , class OutputTypeT , class ParameterVectorType = RealVector>

typedef OutputTypeT shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::OutputType

Defines the output type of the model.

Definition at line 87 of file AbstractModel.h.

Member Enumeration Documentation

◆ Feature

template<class InputTypeT , class OutputTypeT , class ParameterVectorType = RealVector>

enum shark::AbstractModel::Feature

Enumerator
HAS_FIRST_PARAMETER_DERIVATIVE
HAS_FIRST_INPUT_DERIVATIVE

Definition at line 106 of file AbstractModel.h.

Constructor & Destructor Documentation

◆ AbstractModel()

template<class InputTypeT , class OutputTypeT , class ParameterVectorType = RealVector>

shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::AbstractModel ( )

inline

Definition at line 102 of file AbstractModel.h.

◆ ~AbstractModel()

template<class InputTypeT , class OutputTypeT , class ParameterVectorType = RealVector>

virtual shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::~AbstractModel ( )

inlinevirtual

Definition at line 104 of file AbstractModel.h.

Member Function Documentation

◆ createState()

template<class InputTypeT , class OutputTypeT , class ParameterVectorType = RealVector>

virtual boost::shared_ptr< State > shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::createState ( ) const

inlinevirtual

Creates an internal state of the model.

The state is needed when the derivatives are to be calculated. Eval can store a state which is then reused to speed up the calculations of the derivatives. This also allows eval to be evaluated in parallel!

Reimplemented in shark::CMACMap, shark::ConcatenatedModel< VectorType >, shark::Conv2DModel< VectorType, ActivationFunction >, shark::DropoutLayer< VectorType >, shark::KernelExpansion< InputType >, shark::MissingFeaturesKernelExpansion< InputType >, shark::LinearModel< InputType, ActivationFunction >, shark::LinearModel< RealVector >, shark::NeuronLayer< NeuronType, VectorType >, shark::Normalizer< VectorType >, shark::OneVersusOneClassifier< InputType, VectorType >, shark::PoolingLayer< VectorType >, shark::RBFLayer, shark::ResizeLayer< VectorType >, shark::CARTree< LabelType >, and shark::RBM< VisibleLayerT, HiddenLayerT, randomT >.

Definition at line 132 of file AbstractModel.h.

References shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::hasFirstInputDerivative(), shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::hasFirstParameterDerivative(), and SHARKEXCEPTION.

Referenced by shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::eval(), shark::NegativeLogLikelihood::evalDerivative(), and shark::VariationalAutoencoderError< SearchPointType >::evalDerivative().

◆ eval() [1/3]

template<class InputTypeT , class OutputTypeT , class ParameterVectorType = RealVector>

virtual void shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::eval	(	BatchInputType const &	patterns,
		BatchOutputType &	outputs
	)		const

inlinevirtual

Standard interface for evaluating the response of the model to a batch of patterns.

Parameters

patterns	the inputs of the model
outputs	the predictions or response of the model to every pattern

Reimplemented in shark::Normalizer< VectorType >, shark::DropoutLayer< VectorType >, shark::LinearModel< InputType, ActivationFunction >, shark::NeuronLayer< NeuronType, VectorType >, shark::KernelExpansion< InputType >, shark::ConcatenatedModel< VectorType >, shark::MissingFeaturesKernelExpansion< InputType >, shark::LinearModel< InputType, ActivationFunction >, shark::OneVersusOneClassifier< InputType, VectorType >, shark::CARTree< LabelType >, and shark::RBM< VisibleLayerT, HiddenLayerT, randomT >.

Definition at line 163 of file AbstractModel.h.

References shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::createState(), and shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::eval().

Referenced by shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::eval(), shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::eval(), shark::NegativeLogLikelihood::evalDerivative(), shark::VariationalAutoencoderError< SearchPointType >::evalDerivative(), shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::operator()(), and shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::operator()().

◆ eval() [2/3]

template<class InputTypeT , class OutputTypeT , class ParameterVectorType = RealVector>

virtual void shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::eval	(	BatchInputType const &	patterns,
		BatchOutputType &	outputs,
		State &	state
	)		const

pure virtual

Standard interface for evaluating the response of the model to a batch of patterns.

Parameters

patterns	the inputs of the model
outputs	the predictions or response of the model to every pattern
state	intermediate results stored by eval which can be reused for derivative computation.

Implemented in shark::Normalizer< VectorType >, shark::ResizeLayer< VectorType >, shark::Conv2DModel< VectorType, ActivationFunction >, shark::DropoutLayer< VectorType >, shark::LinearModel< InputType, ActivationFunction >, shark::NeuronLayer< NeuronType, VectorType >, shark::PoolingLayer< VectorType >, shark::OneVersusOneClassifier< InputType, VectorType >, shark::ClusteringModel< InputT, OutputT >, shark::ConcatenatedModel< VectorType >, shark::KernelExpansion< InputType >, shark::MissingFeaturesKernelExpansion< InputType >, shark::RBFLayer, shark::CARTree< LabelType >, shark::RBM< VisibleLayerT, HiddenLayerT, randomT >, shark::LinearModel< InputType, ActivationFunction >, and shark::OneVersusOneClassifier< InputType, VectorType >.

◆ eval() [3/3]

template<class InputTypeT , class OutputTypeT , class ParameterVectorType = RealVector>

virtual void shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::eval	(	InputType const &	pattern,
		OutputType &	output
	)		const

inlinevirtual

Standard interface for evaluating the response of the model to a single pattern.

Parameters

pattern	the input of the model
output	the prediction or response of the model to the pattern

Reimplemented in shark::Classifier< Model >, shark::Classifier< detail::BaseNearestNeighbor< InputType, unsigned int > >, shark::Classifier< KernelExpansion< InputType > >, shark::Classifier< LinearModel< RealVector > >, shark::HardClusteringModel< InputT >, shark::SoftClusteringModel< InputT >, shark::LinearModel< InputType, ActivationFunction >, and shark::OneVersusOneClassifier< InputType, VectorType >.

Definition at line 179 of file AbstractModel.h.

References shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::eval(), and shark::getBatchElement().

◆ features()

template<class InputTypeT , class OutputTypeT , class ParameterVectorType = RealVector>

const Features & shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::features ( ) const

inline

Definition at line 110 of file AbstractModel.h.

◆ hasFirstInputDerivative()

template<class InputTypeT , class OutputTypeT , class ParameterVectorType = RealVector>

bool shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::hasFirstInputDerivative ( ) const

inline

Returns true when the first input derivative is implemented.

Definition at line 117 of file AbstractModel.h.

References shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::HAS_FIRST_INPUT_DERIVATIVE, and shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::m_features.

Referenced by shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::createState().

◆ hasFirstParameterDerivative()

template<class InputTypeT , class OutputTypeT , class ParameterVectorType = RealVector>

bool shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::hasFirstParameterDerivative ( ) const

inline

Returns true when the first parameter derivative is implemented.

Definition at line 113 of file AbstractModel.h.

References shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::HAS_FIRST_PARAMETER_DERIVATIVE, and shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::m_features.

Referenced by shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::createState(), shark::NegativeLogLikelihood::NegativeLogLikelihood(), and shark::VariationalAutoencoderError< SearchPointType >::VariationalAutoencoderError().

◆ inputShape()

template<class InputTypeT , class OutputTypeT , class ParameterVectorType = RealVector>

virtual Shape shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::inputShape ( ) const

pure virtual

Returns the expected shape of the input.

Implemented in shark::Classifier< Model >, shark::Classifier< detail::BaseNearestNeighbor< InputType, unsigned int > >, shark::Classifier< KernelExpansion< InputType > >, shark::Classifier< LinearModel< RealVector > >, shark::HardClusteringModel< InputT >, shark::SoftClusteringModel< InputT >, shark::CMACMap, shark::ConcatenatedModel< VectorType >, shark::Conv2DModel< VectorType, ActivationFunction >, shark::DropoutLayer< VectorType >, shark::KernelExpansion< InputType >, shark::LinearModel< InputType, ActivationFunction >, shark::LinearModel< RealVector >, shark::NeuronLayer< NeuronType, VectorType >, shark::Normalizer< VectorType >, shark::OneVersusOneClassifier< InputType, VectorType >, shark::PoolingLayer< VectorType >, shark::RBFLayer, shark::ResizeLayer< VectorType >, shark::CARTree< LabelType >, and shark::RBM< VisibleLayerT, HiddenLayerT, randomT >.

◆ operator()() [1/3]

template<class InputTypeT , class OutputTypeT , class ParameterVectorType = RealVector>

BatchOutputType shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::operator() ( BatchInputType const & patterns ) const

inline

Model evaluation as an operator for a single pattern. This is a convenience function.

Parameters

patterns the input of the model

Returns: the response of the model

Definition at line 209 of file AbstractModel.h.

References shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::eval().

◆ operator()() [2/3]

template<class InputTypeT , class OutputTypeT , class ParameterVectorType = RealVector>

Data< OutputType > shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::operator() ( Data< InputType > const & patterns ) const

inline

Model evaluation as an operator for a whole dataset. This is a convenience function.

Parameters

patterns the input of the model

Returns: the responses of the model

Definition at line 191 of file AbstractModel.h.

References shark::transform().

◆ operator()() [3/3]

template<class InputTypeT , class OutputTypeT , class ParameterVectorType = RealVector>

OutputType shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::operator() ( InputType const & pattern ) const

inline

Model evaluation as an operator for a single pattern. This is a convenience function.

Parameters

pattern the input of the model

Returns: the response of the model

Definition at line 199 of file AbstractModel.h.

References shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::eval().

◆ outputShape()

template<class InputTypeT , class OutputTypeT , class ParameterVectorType = RealVector>

virtual Shape shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::outputShape ( ) const

pure virtual

Returns the shape of the output.

Implemented in shark::Classifier< Model >, shark::Classifier< detail::BaseNearestNeighbor< InputType, unsigned int > >, shark::Classifier< KernelExpansion< InputType > >, shark::Classifier< LinearModel< RealVector > >, shark::HardClusteringModel< InputT >, shark::SoftClusteringModel< InputT >, shark::CMACMap, shark::ConcatenatedModel< VectorType >, shark::Conv2DModel< VectorType, ActivationFunction >, shark::DropoutLayer< VectorType >, shark::KernelExpansion< InputType >, shark::LinearModel< InputType, ActivationFunction >, shark::LinearModel< RealVector >, shark::NeuronLayer< NeuronType, VectorType >, shark::Normalizer< VectorType >, shark::OneVersusOneClassifier< InputType, VectorType >, shark::PoolingLayer< VectorType >, shark::RBFLayer, shark::ResizeLayer< VectorType >, shark::CARTree< LabelType >, and shark::RBM< VisibleLayerT, HiddenLayerT, randomT >.

◆ read()

template<class InputTypeT , class OutputTypeT , class ParameterVectorType = RealVector>

virtual void shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::read ( InArchive & archive )

inlinevirtual

From ISerializable, reads a model from an archive.

Reimplemented from shark::ISerializable.

Reimplemented in shark::Classifier< Model >, shark::Classifier< detail::BaseNearestNeighbor< InputType, unsigned int > >, shark::Classifier< KernelExpansion< InputType > >, shark::Classifier< LinearModel< RealVector > >, shark::ClusteringModel< InputT, OutputT >, shark::ClusteringModel< InputT, RealVector >, shark::ClusteringModel< InputT, unsigned int >, shark::CMACMap, shark::ConcatenatedModel< VectorType >, shark::Conv2DModel< VectorType, ActivationFunction >, shark::DropoutLayer< VectorType >, shark::KernelExpansion< InputType >, shark::LinearModel< InputType, ActivationFunction >, shark::LinearModel< RealVector >, shark::NeuronLayer< NeuronType, VectorType >, shark::Normalizer< VectorType >, shark::OneVersusOneClassifier< InputType, VectorType >, shark::PoolingLayer< VectorType >, shark::RBFLayer, shark::ResizeLayer< VectorType >, shark::CARTree< LabelType >, and shark::RBM< VisibleLayerT, HiddenLayerT, randomT >.

Definition at line 143 of file AbstractModel.h.

References shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::m_features, shark::TypedFlags< Flag >::read(), and shark::IParameterizable< VectorType >::setParameterVector().

◆ updateFeatures()

template<class InputTypeT , class OutputTypeT , class ParameterVectorType = RealVector>

virtual void shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::updateFeatures ( )

inlinevirtual

Definition at line 110 of file AbstractModel.h.

◆ weightedDerivatives()

template<class InputTypeT , class OutputTypeT , class ParameterVectorType = RealVector>

virtual void shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::weightedDerivatives	(	BatchInputType const &	patterns,
		BatchOutputType const &	outputs,
		BatchOutputType const &	coefficients,
		State const &	state,
		ParameterVectorType &	parameterDerivative,
		BatchInputType &	inputDerivative
	)		const

inlinevirtual

calculates weighted input and parameter derivative at the same time

Sometimes, both derivatives are needed at the same time. But sometimes, when calculating the weighted parameter derivative, the input derivative can be calculated for free. This is for example true for the feed-forward neural networks. However, there exists the obvious default implementation that just calculates the derivatives one after another.

Parameters

patterns	the patterns to evaluate
outputs	the target outputs
coefficients	the coefficients which are used to calculate the weighted sum
state	intermediate results stored by eval to sped up calculations of the derivatives
parameterDerivative	the calculated parameter derivative as sum over all derivates of all patterns
inputDerivative	the calculated derivative for every pattern

Reimplemented in shark::ConcatenatedModel< VectorType >.

Definition at line 261 of file AbstractModel.h.

References shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::weightedInputDerivative(), and shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::weightedParameterDerivative().

Referenced by shark::VariationalAutoencoderError< SearchPointType >::evalDerivative().

◆ weightedInputDerivative()

template<class InputTypeT , class OutputTypeT , class ParameterVectorType = RealVector>

virtual void shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::weightedInputDerivative	(	BatchInputType const &	pattern,
		BatchOutputType const &	outputs,
		BatchOutputType const &	coefficients,
		State const &	state,
		BatchInputType &	derivative
	)		const

inlinevirtual

calculates the weighted sum of derivatives w.r.t the inputs

Parameters

pattern	the patterns to evaluate
outputs	the target outputs
coefficients	the coefficients which are used to calculate the weighted sum for every pattern
state	intermediate results stored by eval to sped up calculations of the derivatives
derivative	the calculated derivative for every pattern

Reimplemented in shark::NeuronLayer< NeuronType, VectorType >, shark::PoolingLayer< VectorType >, shark::Conv2DModel< VectorType, ActivationFunction >, shark::ResizeLayer< VectorType >, and shark::DropoutLayer< VectorType >.

Definition at line 239 of file AbstractModel.h.

References shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::HAS_FIRST_INPUT_DERIVATIVE, and SHARK_FEATURE_EXCEPTION.

Referenced by shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::weightedDerivatives().

◆ weightedParameterDerivative()

template<class InputTypeT , class OutputTypeT , class ParameterVectorType = RealVector>

virtual void shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::weightedParameterDerivative	(	BatchInputType const &	pattern,
		BatchOutputType const &	outputs,
		BatchOutputType const &	coefficients,
		State const &	state,
		ParameterVectorType &	derivative
	)		const

inlinevirtual

calculates the weighted sum of derivatives w.r.t the parameters.

Parameters

pattern	the patterns to evaluate
outputs	the target outputs
coefficients	the coefficients which are used to calculate the weighted sum for every pattern
state	intermediate results stored by eval to speed up calculations of the derivatives
derivative	the calculated derivative as sum over all derivates of all patterns

Reimplemented in shark::Conv2DModel< VectorType, ActivationFunction >, shark::NeuronLayer< NeuronType, VectorType >, shark::PoolingLayer< VectorType >, shark::ResizeLayer< VectorType >, shark::RBFLayer, shark::ConcatenatedModel< VectorType >, and shark::DropoutLayer< VectorType >.

Definition at line 222 of file AbstractModel.h.

References shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::HAS_FIRST_PARAMETER_DERIVATIVE, and SHARK_FEATURE_EXCEPTION.

Referenced by shark::NegativeLogLikelihood::evalDerivative(), shark::VariationalAutoencoderError< SearchPointType >::evalDerivative(), and shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::weightedDerivatives().

◆ write()

template<class InputTypeT , class OutputTypeT , class ParameterVectorType = RealVector>

virtual void shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::write ( OutArchive & archive ) const

inlinevirtual

writes a model to an archive

the default implementation just saves the parameters, not the structure!

Reimplemented from shark::ISerializable.

Reimplemented in shark::Classifier< Model >, shark::Classifier< detail::BaseNearestNeighbor< InputType, unsigned int > >, shark::Classifier< KernelExpansion< InputType > >, shark::Classifier< LinearModel< RealVector > >, shark::ClusteringModel< InputT, OutputT >, shark::ClusteringModel< InputT, RealVector >, shark::ClusteringModel< InputT, unsigned int >, shark::CMACMap, shark::ConcatenatedModel< VectorType >, shark::Conv2DModel< VectorType, ActivationFunction >, shark::DropoutLayer< VectorType >, shark::KernelExpansion< InputType >, shark::LinearModel< InputType, ActivationFunction >, shark::LinearModel< RealVector >, shark::NeuronLayer< NeuronType, VectorType >, shark::Normalizer< VectorType >, shark::OneVersusOneClassifier< InputType, VectorType >, shark::PoolingLayer< VectorType >, shark::RBFLayer, shark::ResizeLayer< VectorType >, shark::CARTree< LabelType >, and shark::RBM< VisibleLayerT, HiddenLayerT, randomT >.

Definition at line 153 of file AbstractModel.h.

References shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::m_features, shark::IParameterizable< VectorType >::parameterVector(), and shark::TypedFlags< Flag >::write().

Member Data Documentation

◆ m_features

template<class InputTypeT , class OutputTypeT , class ParameterVectorType = RealVector>

Features shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::m_features

protected

Definition at line 110 of file AbstractModel.h.

Referenced by shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::hasFirstInputDerivative(), shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::hasFirstParameterDerivative(), shark::LinearModel< InputType, ActivationFunction >::LinearModel(), shark::LinearModel< InputType, ActivationFunction >::LinearModel(), shark::LinearModel< InputType, ActivationFunction >::LinearModel(), shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::read(), and shark::AbstractModel< InputTypeT, OutputTypeT, ParameterVectorType >::write().

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

include/shark/Models/AbstractModel.h

Public Types

Public Member Functions

Protected Attributes

Detailed Description

Member Typedef Documentation

◆ BatchInputType

◆ BatchOutputType

◆ FeatureNotAvailableException

◆ Features

◆ InputType

◆ ModelBaseType

◆ OutputType

Member Enumeration Documentation

◆ Feature

Constructor & Destructor Documentation

◆ AbstractModel()

◆ ~AbstractModel()

Member Function Documentation

◆ createState()

◆ eval() [1/3]

◆ eval() [2/3]

◆ eval() [3/3]

◆ features()

◆ hasFirstInputDerivative()

◆ hasFirstParameterDerivative()

◆ inputShape()

◆ operator()() [1/3]

◆ operator()() [2/3]

◆ operator()() [3/3]

◆ outputShape()

◆ read()

◆ updateFeatures()

◆ weightedDerivatives()

◆ weightedInputDerivative()

◆ weightedParameterDerivative()

◆ write()

Member Data Documentation

◆ m_features