Model which corrupts the data using gaussian noise. More...

#include <shark/Models/GaussianNoiseModel.h>

Inheritance diagram for shark::GaussianNoiseModel:

Public Member Functions
	GaussianNoiseModel ()
	Default Constructor; use setStructure later. More...

	GaussianNoiseModel (unsigned int inputs, double variance)
	Constructor creating a model with given input size and the same variance for all inputs. More...

std::string	name () const
	From INameable: return the class name. More...

size_t	inputSize () const
	obtain the input dimension More...

size_t	outputSize () const
	obtain the output dimension More...

RealVector	parameterVector () const
	obtain the parameter vector More...

void	setParameterVector (RealVector const &newParameters)
	overwrite the parameter vector More...

size_t	numberOfParameters () const
	return the number of parameter More...

void	setStructure (unsigned int inputs, double variance)
	overwrite structure and parameters More...

void	setStructure (RealVector const &variances)
	overwrite structure and parameters More...

RealVector const &	variances () const

RealVector &	variances ()

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

void	eval (BatchInputType const &inputs, BatchOutputType &outputs) const
	Add noise to the input. More...

void	eval (BatchInputType const &inputs, BatchOutputType &outputs, State &state) const
	Evaluate the model: output = matrix * input + offset. More...

void	weightedParameterDerivative (BatchInputType const &patterns, RealVector const &coefficients, State const &state, RealVector &gradient) const

Public Member Functions inherited from shark::AbstractModel< RealVector, RealVector >
	AbstractModel ()

virtual	~AbstractModel ()

const Features &	features () const

virtual void	updateFeatures ()

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

bool	hasSecondParameterDerivative () const
	Returns true when the second parameter derivative is implemented. More...

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

bool	hasSecondInputDerivative () const
	Returns true when the second parameter derivative is implemented. More...

bool	isSequential () const

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

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

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

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

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

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

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

virtual void	weightedParameterDerivative (BatchInputType const &pattern, BatchOutputType const &coefficients, Batch< RealMatrix >::type const &errorHessian, State const &state, RealVector &derivative, RealMatrix &hessian) const
	calculates the weighted sum of derivatives w.r.t the parameters More...

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

virtual void	weightedInputDerivative (BatchInputType const &pattern, BatchOutputType const &coefficients, typename Batch< RealMatrix >::type const &errorHessian, State const &state, RealMatrix &derivative, Batch< RealMatrix >::type &hessian) const
	calculates the weighted sum of derivatives w.r.t the inputs More...

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

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

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

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 ()

Additional Inherited Members
Public Types inherited from shark::AbstractModel< RealVector, RealVector >
enum	Feature

typedef RealVector	InputType
	Defines the input type of the model. More...

typedef RealVector	OutputType
	Defines the output type of the model. More...

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

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

typedef TypedFlags< Feature >	Features

typedef TypedFeatureNotAvailableException< Feature >	FeatureNotAvailableException

Protected Attributes inherited from shark::AbstractModel< RealVector, RealVector >
Features	m_features

Detailed Description

Model which corrupts the data using gaussian noise.

When training autoencoders, it proved beneficial to add noise to the input and train the model to remove that noise again, instead of only larning a identity transformation. This Model represents one choice of Noise: Gaussian Noise, to do this. the formula of corruption of an input \( x=(x_1,\dots,x_n) \) with variances \( \sigma = (\sigma_1, \dots, \sigma_n) \) is

\[ x_i \leftarrow x_i + N(0,\sigma_i) \]

Usage is simple. given your autoencoder/decoder pair ConvatenatedModel<RealVector,RealVector> autoencoder = encoder >> decoder; we can just concatenate this model: GaussianNoiseModel noise(0.1);//variance of noise ConvatenatedModel<RealVector,RealVector> denoisingAutoencoder = noise>>autoencoder; and train the model using the standard autoencoder error

Definition at line 55 of file GaussianNoiseModel.h.

Constructor & Destructor Documentation

◆ GaussianNoiseModel() [1/2]

shark::GaussianNoiseModel::GaussianNoiseModel ( )

inline

Default Constructor; use setStructure later.

Definition at line 63 of file GaussianNoiseModel.h.

References shark::AbstractModel< RealVector, RealVector >::HAS_FIRST_PARAMETER_DERIVATIVE, and shark::AbstractModel< RealVector, RealVector >::m_features.

◆ GaussianNoiseModel() [2/2]

shark::GaussianNoiseModel::GaussianNoiseModel	(	unsigned int	inputs,
		double	variance
	)

inline

Constructor creating a model with given input size and the same variance for all inputs.

Definition at line 67 of file GaussianNoiseModel.h.

References shark::AbstractModel< RealVector, RealVector >::HAS_FIRST_PARAMETER_DERIVATIVE, and shark::AbstractModel< RealVector, RealVector >::m_features.

Member Function Documentation

◆ createState()

boost::shared_ptr<State> shark::GaussianNoiseModel::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 from shark::AbstractModel< RealVector, RealVector >.

Definition at line 120 of file GaussianNoiseModel.h.

◆ eval() [1/2]

void shark::GaussianNoiseModel::eval	(	BatchInputType const &	inputs,
		BatchOutputType &	outputs
	)		const

inlinevirtual

Add noise to the input.

Reimplemented from shark::AbstractModel< RealVector, RealVector >.

Definition at line 125 of file GaussianNoiseModel.h.

References shark::gauss(), inputSize(), SHARK_CRITICAL_REGION, and SIZE_CHECK.

Referenced by eval().

◆ eval() [2/2]

void shark::GaussianNoiseModel::eval	(	BatchInputType const &	inputs,
		BatchOutputType &	outputs,
		State &	state
	)		const

inlinevirtual

Evaluate the model: output = matrix * input + offset.

Implements shark::AbstractModel< RealVector, RealVector >.

Definition at line 141 of file GaussianNoiseModel.h.

References eval().

◆ inputSize()

size_t shark::GaussianNoiseModel::inputSize ( ) const

inline

obtain the input dimension

Definition at line 77 of file GaussianNoiseModel.h.

Referenced by eval().

◆ name()

std::string shark::GaussianNoiseModel::name ( ) const

inlinevirtual

From INameable: return the class name.

Reimplemented from shark::INameable.

Definition at line 73 of file GaussianNoiseModel.h.

◆ numberOfParameters()

size_t shark::GaussianNoiseModel::numberOfParameters ( ) const

inlinevirtual

return the number of parameter

Reimplemented from shark::IParameterizable.

Definition at line 98 of file GaussianNoiseModel.h.

◆ outputSize()

size_t shark::GaussianNoiseModel::outputSize ( ) const

inline

obtain the output dimension

Definition at line 82 of file GaussianNoiseModel.h.

◆ parameterVector()

RealVector shark::GaussianNoiseModel::parameterVector ( ) const

inlinevirtual

obtain the parameter vector

Reimplemented from shark::IParameterizable.

Definition at line 87 of file GaussianNoiseModel.h.

◆ setParameterVector()

void shark::GaussianNoiseModel::setParameterVector ( RealVector const & newParameters )

inlinevirtual

overwrite the parameter vector

Reimplemented from shark::IParameterizable.

Definition at line 92 of file GaussianNoiseModel.h.

References SIZE_CHECK.

◆ setStructure() [1/2]

void shark::GaussianNoiseModel::setStructure	(	unsigned int	inputs,
		double	variance
	)

inline

overwrite structure and parameters

Definition at line 103 of file GaussianNoiseModel.h.

◆ setStructure() [2/2]

void shark::GaussianNoiseModel::setStructure ( RealVector const & variances )

inline

overwrite structure and parameters

Definition at line 108 of file GaussianNoiseModel.h.

References variances().

◆ variances() [1/2]

RealVector const& shark::GaussianNoiseModel::variances ( ) const

inline

Definition at line 112 of file GaussianNoiseModel.h.

Referenced by setStructure().

◆ variances() [2/2]

RealVector& shark::GaussianNoiseModel::variances ( )

inline

Definition at line 116 of file GaussianNoiseModel.h.

◆ weightedParameterDerivative()

void shark::GaussianNoiseModel::weightedParameterDerivative	(	BatchInputType const &	patterns,
		RealVector const &	coefficients,
		State const &	state,
		RealVector &	gradient
	)		const

inline

Definition at line 145 of file GaussianNoiseModel.h.

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

include/shark/Models/GaussianNoiseModel.h

Public Member Functions

Additional Inherited Members

Detailed Description

Constructor & Destructor Documentation

◆ GaussianNoiseModel() [1/2]

◆ GaussianNoiseModel() [2/2]

Member Function Documentation

◆ createState()

◆ eval() [1/2]

◆ eval() [2/2]

◆ inputSize()

◆ name()

◆ numberOfParameters()

◆ outputSize()

◆ parameterVector()

◆ setParameterVector()

◆ setStructure() [1/2]

◆ setStructure() [2/2]

◆ variances() [1/2]

◆ variances() [2/2]

◆ weightedParameterDerivative()