Limited-Memory Broyden, Fletcher, Goldfarb, Shannon algorithm for unconstrained optimization. More...

#include <shark/Algorithms/GradientDescent/LBFGS.h>

Inheritance diagram for shark::LBFGS:

Public Member Functions
	LBFGS ()

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

void	setHistCount (unsigned int numhist)
	Specify the amount of steps to be memorized and used to find the L-BFGS direction. More...

SHARK_EXPORT_SYMBOL void	read (InArchive &archive)
	Read the component from the supplied archive. More...

SHARK_EXPORT_SYMBOL void	write (OutArchive &archive) const
	Write the component to the supplied archive. More...

Public Member Functions inherited from shark::AbstractLineSearchOptimizer
SHARK_EXPORT_SYMBOL	AbstractLineSearchOptimizer ()

SHARK_EXPORT_SYMBOL void	init (ObjectiveFunctionType &objectiveFunction, SearchPointType const &startingPoint)
	initializes the optimizer using a predefined starting point More...

SHARK_EXPORT_SYMBOL void	step (ObjectiveFunctionType const &objectiveFunction)
	Carry out one step of the optimizer for the supplied objective function. More...

LineSearch const &	lineSearch () const

LineSearch &	lineSearch ()

Public Member Functions inherited from shark::AbstractSingleObjectiveOptimizer< RealVector >
virtual void	init (ObjectiveFunctionType &function)

virtual const SolutionType &	solution () const
	returns the current solution of the optimizer More...

Public Member Functions inherited from shark::AbstractOptimizer< RealVector, double, SingleObjectiveResultSet< RealVector > >
const Features &	features () const

virtual void	updateFeatures ()

bool	requiresValue () const

bool	requiresFirstDerivative () const

bool	requiresSecondDerivative () const

bool	canSolveConstrained () const

bool	requiresClosestFeasible () const

virtual	~AbstractOptimizer ()
	Empty virtual d'tor. More...

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 ()

Protected Member Functions
SHARK_EXPORT_SYMBOL void	initModel ()
	Initializes the internal model. More...

SHARK_EXPORT_SYMBOL void	computeSearchDirection ()
	Updates the Model and computes the next search direction. More...

SHARK_EXPORT_SYMBOL void	updateHist (RealVector &y, RealVector &step)
	Stores another step and searchDirection, discarding the oldest on if necessary. More...

SHARK_EXPORT_SYMBOL void	getDirection (RealVector &searchDirection)
	Get the LBFGS direction. More...

Protected Member Functions inherited from shark::AbstractOptimizer< RealVector, double, SingleObjectiveResultSet< RealVector > >
void	checkFeatures (ObjectiveFunctionType const &objectiveFunction)
	Convenience function that checks whether the features of the supplied objective function match with the required features of the optimizer. More...

Protected Attributes
double	m_updThres
	Threshold for when to update history. More...

unsigned int	m_numHist
	Number of steps to use for LBFGS. More...

double	m_hdiag

std::deque< RealVector >	m_steps

std::deque< RealVector >	m_gradientDifferences

Protected Attributes inherited from shark::AbstractLineSearchOptimizer
LineSearch	m_linesearch
	used line search method. More...

std::size_t	m_dimension
	number of parameters More...

double	m_initialStepLength
	Initial step length to begin with the line search. More...

RealVector	m_derivative
	gradient of m_best.point More...

RealVector	m_searchDirection
	search direction of next step More...

RealVector	m_lastPoint
	previous point More...

RealVector	m_lastDerivative
	gradient of the previous point More...

double	m_lastValue
	value of the previous point More...

Protected Attributes inherited from shark::AbstractSingleObjectiveOptimizer< RealVector >
SolutionType	m_best
	current solution of the optimizer More...

Protected Attributes inherited from shark::AbstractOptimizer< RealVector, double, SingleObjectiveResultSet< RealVector > >
Features	m_features

Additional Inherited Members
Public Types inherited from shark::AbstractSingleObjectiveOptimizer< RealVector >
typedef base_type::SearchPointType	SearchPointType

typedef base_type::SolutionType	SolutionType

typedef base_type::ResultType	ResultType

typedef base_type::ObjectiveFunctionType	ObjectiveFunctionType

Public Types inherited from shark::AbstractOptimizer< RealVector, double, SingleObjectiveResultSet< RealVector > >
enum	Feature
	Models features that the optimizer requires from the objective function. More...

typedef RealVector	SearchPointType

typedef double	ResultType

typedef SingleObjectiveResultSet< RealVector >	SolutionType

typedef AbstractObjectiveFunction< RealVector, ResultType >	ObjectiveFunctionType

typedef TypedFlags< Feature >	Features

typedef TypedFeatureNotAvailableException< Feature >	FeatureNotAvailableException

Detailed Description

Limited-Memory Broyden, Fletcher, Goldfarb, Shannon algorithm for unconstrained optimization.

Definition at line 50 of file LBFGS.h.

Constructor & Destructor Documentation

§ LBFGS()

shark::LBFGS::LBFGS ( )

inline

Definition at line 55 of file LBFGS.h.

Member Function Documentation

§ computeSearchDirection()

SHARK_EXPORT_SYMBOL void shark::LBFGS::computeSearchDirection ( )

protectedvirtual

Updates the Model and computes the next search direction.

After a step was performed, this method is called to compute the next search direction. This usually involves updating the internal model using the new and old step information. Afterwards m_searchDirection should contain the next search direction.

Implements shark::AbstractLineSearchOptimizer.

§ getDirection()

SHARK_EXPORT_SYMBOL void shark::LBFGS::getDirection ( RealVector & searchDirection )

protected

Get the LBFGS direction.

This approximates the inverse hessian multiplied by the gradient. This uses the rho, alpha and beta vectors. Description of these can be seen in ie. the wiki page of LBFGS.

Referenced by setHistCount().

§ initModel()

SHARK_EXPORT_SYMBOL void shark::LBFGS::initModel ( )

protectedvirtual

Initializes the internal model.

Line Search Methods use a Model to search for the next search direction. The model is initialized during init()

Implements shark::AbstractLineSearchOptimizer.

§ name()

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

inlinevirtual

From INameable: return the class name.

Reimplemented from shark::INameable.

Definition at line 58 of file LBFGS.h.

§ read()

SHARK_EXPORT_SYMBOL void shark::LBFGS::read ( InArchive & archive )

virtual

Read the component from the supplied archive.

Parameters

[in,out] archive The archive to read from.

Reimplemented from shark::AbstractLineSearchOptimizer.

Referenced by setHistCount().

§ setHistCount()

void shark::LBFGS::setHistCount ( unsigned int numhist )

inline

Specify the amount of steps to be memorized and used to find the L-BFGS direction.

Parameters

numhist The amount of steps to use.

Definition at line 64 of file LBFGS.h.

References getDirection(), m_numHist, read(), SHARK_CHECK, SHARK_EXPORT_SYMBOL, shark::AbstractLineSearchOptimizer::step(), updateHist(), and write().

§ updateHist()

SHARK_EXPORT_SYMBOL void shark::LBFGS::updateHist	(	RealVector &	y,
		RealVector &	step
	)

protected

Stores another step and searchDirection, discarding the oldest on if necessary.

Parameters

step	Last performed step
y	difference in gradients

Referenced by setHistCount().

§ write()

SHARK_EXPORT_SYMBOL void shark::LBFGS::write ( OutArchive & archive ) const

virtual

Write the component to the supplied archive.

Parameters

[in,out] archive The archive to write to.

Reimplemented from shark::AbstractLineSearchOptimizer.

Referenced by setHistCount().

Member Data Documentation

§ m_gradientDifferences

std::deque<RealVector> shark::LBFGS::m_gradientDifferences

protected

Definition at line 99 of file LBFGS.h.

§ m_hdiag

double shark::LBFGS::m_hdiag

protected

Definition at line 92 of file LBFGS.h.

§ m_numHist

unsigned int shark::LBFGS::m_numHist

protected

Number of steps to use for LBFGS.

Definition at line 89 of file LBFGS.h.

Referenced by setHistCount().

§ m_steps

std::deque<RealVector> shark::LBFGS::m_steps

protected

Definition at line 98 of file LBFGS.h.

§ m_updThres

double shark::LBFGS::m_updThres

protected

Threshold for when to update history.

Definition at line 88 of file LBFGS.h.

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

include/shark/Algorithms/GradientDescent/LBFGS.h

Public Member Functions

Protected Member Functions

Protected Attributes

Additional Inherited Members

Detailed Description

Constructor & Destructor Documentation

§ LBFGS()

Member Function Documentation

§ computeSearchDirection()

§ getDirection()

§ initModel()

§ name()

§ read()

§ setHistCount()

§ updateHist()

§ write()

Member Data Documentation

§ m_gradientDifferences

§ m_hdiag

§ m_numHist

§ m_steps

§ m_updThres