#include <itkStandardStochasticGradientDescentOptimizer.h>

Detailed Description

This class implements a gradient descent optimizer with a decaying gain.

If is a costfunction that has to be minimised, the following iterative algorithm is used to find the optimal parameters :

$\‍[ x(k+1) = x(k) - a(k) dC/dx \‍]$

The gain at each iteration is defined by:

$\‍[ a(k) = a / (A + k + 1)^alpha \‍]$

.

It is very suitable to be used in combination with a stochastic estimate of the gradient . For example, in image registration problems it is often advantageous to compute the metric derivative ( ) on a new set of randomly selected image samples in each iteration. You may set the parameter NewSamplesEveryIteration to "true" to achieve this effect. For more information on this strategy, you may have a look at:

S. Klein, M. Staring, J.P.W. Pluim, "Comparison of gradient approximation techniques for optimisation of mutual information in nonrigid registration", in: SPIE Medical Imaging: Image Processing, Editor(s): J.M. Fitzpatrick, J.M. Reinhardt, SPIE press, 2005, vol. 5747, Proceedings of SPIE, pp. 192-203.

Or:

S. Klein, M. Staring, J.P.W. Pluim, "Evaluation of Optimization Methods for Nonrigid Medical Image Registration using Mutual Information and B-Splines" IEEE Transactions on Image Processing, 2007, nr. 16(12), December.

This class also serves as a base class for other StochasticGradient type algorithms, like the AcceleratedStochasticGradientOptimizer.

See also: StandardStochasticGradient, AcceleratedStochasticGradientOptimizer

Definition at line 64 of file itkStandardStochasticGradientDescentOptimizer.h.

Inheritance diagram for itk::StandardStochasticGradientOptimizer:

Public Types
using	ConstPointer = SmartPointer<const Self>
using	Pointer = SmartPointer<Self>
using	Self = StandardStochasticGradientOptimizer
enum	StopConditionType
using	Superclass = StochasticGradientDescentOptimizer
Public Types inherited from itk::StochasticGradientDescentOptimizer
using	ConstPointer = SmartPointer<const Self>
using	Pointer = SmartPointer<Self>
using	ScaledCostFunctionPointer
using	ScaledCostFunctionType
using	ScalesType
using	Self = StochasticGradientDescentOptimizer
enum	StopConditionType { MaximumNumberOfIterations , MetricError , MinimumStepSize , InvalidDiagonalMatrix , GradientMagnitudeTolerance , LineSearchError }
using	Superclass = ScaledSingleValuedNonLinearOptimizer
Public Types inherited from itk::ScaledSingleValuedNonLinearOptimizer
using	ConstPointer = SmartPointer<const Self>
using	Pointer = SmartPointer<Self>
using	ScaledCostFunctionPointer = ScaledCostFunctionType::Pointer
using	ScaledCostFunctionType = ScaledSingleValuedCostFunction
using	ScalesType = NonLinearOptimizer::ScalesType
using	Self = ScaledSingleValuedNonLinearOptimizer
using	Superclass = SingleValuedNonLinearOptimizer

Public Member Functions
void	AdvanceOneStep () override
virtual double	GetCurrentTime () const
virtual double	GetInitialTime () const
virtual double	GetParam_A () const
virtual double	GetParam_a () const
virtual double	GetParam_alpha () const
virtual double	GetParam_beta () const
	ITK_DISALLOW_COPY_AND_MOVE (StandardStochasticGradientOptimizer)
	itkOverrideGetNameOfClassMacro (StandardStochasticGradientOptimizer)
virtual void	ResetCurrentTimeToInitialTime ()
virtual void	SetInitialTime (double _arg)
virtual void	SetParam_A (double _arg)
virtual void	SetParam_a (double _arg)
virtual void	SetParam_alpha (double _arg)
virtual void	SetParam_beta (double _arg)
void	StartOptimization () override
Public Member Functions inherited from itk::StochasticGradientDescentOptimizer
virtual unsigned int	GetCurrentInnerIteration () const
virtual unsigned int	GetCurrentIteration () const
virtual const DerivativeType &	GetGradient ()
virtual unsigned int	GetLBFGSMemory () const
virtual const double &	GetLearningRate ()
virtual const unsigned long &	GetNumberOfInnerIterations ()
virtual const unsigned long &	GetNumberOfIterations ()
virtual const DerivativeType &	GetPreviousGradient ()
virtual const ParametersType &	GetPreviousPosition ()
virtual const DerivativeType &	GetSearchDir ()
virtual const StopConditionType &	GetStopCondition ()
virtual const double &	GetValue ()
	ITK_DISALLOW_COPY_AND_MOVE (StochasticGradientDescentOptimizer)
	itkOverrideGetNameOfClassMacro (StochasticGradientDescentOptimizer)
virtual void	MetricErrorResponse (ExceptionObject &err)
virtual void	ResumeOptimization ()
virtual void	SetLearningRate (double _arg)
virtual void	SetNumberOfIterations (unsigned long _arg)
void	SetNumberOfWorkUnits (ThreadIdType numberOfThreads)
virtual void	SetPreviousGradient (DerivativeType _arg)
virtual void	SetPreviousPosition (ParametersType _arg)
virtual void	SetUseEigen (bool _arg)
virtual void	SetUseMultiThread (bool _arg)
void	StartOptimization () override
virtual void	StopOptimization ()
Public Member Functions inherited from itk::ScaledSingleValuedNonLinearOptimizer
const ParametersType &	GetCurrentPosition () const override
virtual bool	GetMaximize () const
virtual const ScaledCostFunctionType *	GetScaledCostFunction ()
virtual const ParametersType &	GetScaledCurrentPosition ()
bool	GetUseScales () const
virtual void	InitializeScales ()
	ITK_DISALLOW_COPY_AND_MOVE (ScaledSingleValuedNonLinearOptimizer)
	itkOverrideGetNameOfClassMacro (ScaledSingleValuedNonLinearOptimizer)
virtual void	MaximizeOff ()
virtual void	MaximizeOn ()
void	SetCostFunction (CostFunctionType *costFunction) override
virtual void	SetMaximize (bool _arg)
virtual void	SetUseScales (bool arg)

Static Public Member Functions
static Pointer	New ()
Static Public Member Functions inherited from itk::StochasticGradientDescentOptimizer
static Pointer	New ()
Static Public Member Functions inherited from itk::ScaledSingleValuedNonLinearOptimizer
static Pointer	New ()

Protected Member Functions
virtual double	Compute_a (double k) const
virtual double	Compute_beta (double k) const
	StandardStochasticGradientOptimizer ()
virtual void	UpdateCurrentTime ()
	~StandardStochasticGradientOptimizer () override=default
Protected Member Functions inherited from itk::StochasticGradientDescentOptimizer
void	PrintSelf (std::ostream &os, Indent indent) const override
	StochasticGradientDescentOptimizer ()
	~StochasticGradientDescentOptimizer () override=default
Protected Member Functions inherited from itk::ScaledSingleValuedNonLinearOptimizer
virtual void	GetScaledDerivative (const ParametersType &parameters, DerivativeType &derivative) const
virtual MeasureType	GetScaledValue (const ParametersType &parameters) const
virtual void	GetScaledValueAndDerivative (const ParametersType &parameters, MeasureType &value, DerivativeType &derivative) const
void	PrintSelf (std::ostream &os, Indent indent) const override
	ScaledSingleValuedNonLinearOptimizer ()
void	SetCurrentPosition (const ParametersType &param) override
virtual void	SetScaledCurrentPosition (const ParametersType &parameters)
	~ScaledSingleValuedNonLinearOptimizer () override=default

Protected Attributes
double	m_CurrentTime { 0.0 }
bool	m_UseConstantStep {}
Protected Attributes inherited from itk::StochasticGradientDescentOptimizer
unsigned long	m_CurrentInnerIteration {}
unsigned long	m_CurrentIteration { 0 }
DerivativeType	m_Gradient {}
unsigned long	m_LBFGSMemory { 0 }
double	m_LearningRate { 1.0 }
ParametersType	m_MeanSearchDir {}
unsigned long	m_NumberOfInnerIterations {}
unsigned long	m_NumberOfIterations { 100 }
DerivativeType	m_PrePreviousGradient {}
ParametersType	m_PrePreviousSearchDir {}
DerivativeType	m_PreviousGradient {}
ParametersType	m_PreviousPosition {}
ParametersType	m_PreviousSearchDir {}
ParametersType	m_SearchDir {}
bool	m_Stop { false }
StopConditionType	m_StopCondition { MaximumNumberOfIterations }
MultiThreaderBase::Pointer	m_Threader { MultiThreaderBase::New() }
double	m_Value { 0.0 }
Protected Attributes inherited from itk::ScaledSingleValuedNonLinearOptimizer
ScaledCostFunctionPointer	m_ScaledCostFunction {}
ParametersType	m_ScaledCurrentPosition {}

Private Attributes
double	m_InitialTime { 0.0 }
double	m_Param_A { 1.0 }
double	m_Param_a { 1.0 }
double	m_Param_alpha { 0.602 }
double	m_Param_beta {}

Additional Inherited Members
Protected Types inherited from itk::StochasticGradientDescentOptimizer
using	ThreadInfoType = MultiThreaderBase::WorkUnitInfo

Member Typedef Documentation

◆ ConstPointer

using itk::StandardStochasticGradientOptimizer::ConstPointer = SmartPointer<const Self>

Definition at line 73 of file itkStandardStochasticGradientDescentOptimizer.h.

◆ Pointer

using itk::StandardStochasticGradientOptimizer::Pointer = SmartPointer<Self>

Definition at line 72 of file itkStandardStochasticGradientDescentOptimizer.h.

◆ Self

using itk::StandardStochasticGradientOptimizer::Self = StandardStochasticGradientOptimizer

Standard ITK.

Definition at line 70 of file itkStandardStochasticGradientDescentOptimizer.h.

◆ Superclass

using itk::StandardStochasticGradientOptimizer::Superclass = StochasticGradientDescentOptimizer

Definition at line 71 of file itkStandardStochasticGradientDescentOptimizer.h.

Member Enumeration Documentation

◆ StopConditionType

enum itk::StochasticGradientDescentOptimizer::StopConditionType

Codes of stopping conditions The MinimumStepSize stopcondition never occurs, but may be implemented in inheriting classes

Definition at line 82 of file itkStochasticGradientDescentOptimizer.h.

Constructor & Destructor Documentation

◆ StandardStochasticGradientOptimizer()

itk::StandardStochasticGradientOptimizer::StandardStochasticGradientOptimizer ( )

protected

◆ ~StandardStochasticGradientOptimizer()

itk::StandardStochasticGradientOptimizer::~StandardStochasticGradientOptimizer ( )

overrideprotecteddefault

Member Function Documentation

◆ AdvanceOneStep()

void itk::StandardStochasticGradientOptimizer::AdvanceOneStep ( )

overridevirtual

Sets a new LearningRate before calling the Superclass' implementation, and updates the current time.

Reimplemented from itk::StochasticGradientDescentOptimizer.

◆ Compute_a()

virtual double itk::StandardStochasticGradientOptimizer::Compute_a ( double k ) const

protectedvirtual

Function to compute the step size for SGD at time/iteration k.

◆ Compute_beta()

virtual double itk::StandardStochasticGradientOptimizer::Compute_beta ( double k ) const

protectedvirtual

Function to compute the step size for SQN at time/iteration k.

◆ GetCurrentTime()

virtual double itk::StandardStochasticGradientOptimizer::GetCurrentTime ( ) const

virtual

Get the current time. This equals the CurrentIteration in this base class but may be different in inheriting classes, such as the AccelerateStochasticGradient

◆ GetInitialTime()

virtual double itk::StandardStochasticGradientOptimizer::GetInitialTime ( ) const

virtual

◆ GetParam_A()

virtual double itk::StandardStochasticGradientOptimizer::GetParam_A ( ) const

virtual

◆ GetParam_a()

virtual double itk::StandardStochasticGradientOptimizer::GetParam_a ( ) const

virtual

◆ GetParam_alpha()

virtual double itk::StandardStochasticGradientOptimizer::GetParam_alpha ( ) const

virtual

◆ GetParam_beta()

virtual double itk::StandardStochasticGradientOptimizer::GetParam_beta ( ) const

virtual

◆ ITK_DISALLOW_COPY_AND_MOVE()

itk::StandardStochasticGradientOptimizer::ITK_DISALLOW_COPY_AND_MOVE ( StandardStochasticGradientOptimizer )

◆ itkOverrideGetNameOfClassMacro()

itk::StandardStochasticGradientOptimizer::itkOverrideGetNameOfClassMacro ( StandardStochasticGradientOptimizer )

Run-time type information (and related methods).

◆ New()

Pointer itk::StandardStochasticGradientOptimizer::New ( )

static

Method for creation through the object factory.

◆ ResetCurrentTimeToInitialTime()

virtual void itk::StandardStochasticGradientOptimizer::ResetCurrentTimeToInitialTime ( )

inlinevirtual

Set the current time to the initial time. This can be useful to 'reset' the optimisation, for example if you changed the cost function while optimisation. Be careful with this function.

Definition at line 132 of file itkStandardStochasticGradientDescentOptimizer.h.

◆ SetInitialTime()

virtual void itk::StandardStochasticGradientOptimizer::SetInitialTime ( double _arg )

virtual

Set/Get the initial time. Should be >=0. This function is superfluous, since Param_A does effectively the same. However, in inheriting classes, like the AcceleratedStochasticGradient the initial time may have a different function than Param_A. Default: 0.0

◆ SetParam_A()

virtual void itk::StandardStochasticGradientOptimizer::SetParam_A ( double _arg )

virtual

Set/Get A.

◆ SetParam_a()

virtual void itk::StandardStochasticGradientOptimizer::SetParam_a ( double _arg )

virtual

Set/Get a.

◆ SetParam_alpha()

virtual void itk::StandardStochasticGradientOptimizer::SetParam_alpha ( double _arg )

virtual

Set/Get alpha.

◆ SetParam_beta()

virtual void itk::StandardStochasticGradientOptimizer::SetParam_beta ( double _arg )

virtual

Set/Get beta.

◆ StartOptimization()

void itk::StandardStochasticGradientOptimizer::StartOptimization ( )

override

Set current time to 0 and call superclass' implementation.

◆ UpdateCurrentTime()

virtual void itk::StandardStochasticGradientOptimizer::UpdateCurrentTime ( )

protectedvirtual

Function to update the current time This function just increments the CurrentTime by 1. Inheriting functions may implement something smarter, for example, dependent on the progress

Reimplemented in itk::AdaptiveStochasticLBFGSOptimizer.

Member Data Documentation

◆ m_CurrentTime

double itk::StandardStochasticGradientOptimizer::m_CurrentTime { 0.0 }

protected

The current time, which serves as input for Compute_a

Definition at line 157 of file itkStandardStochasticGradientDescentOptimizer.h.

◆ m_InitialTime

double itk::StandardStochasticGradientOptimizer::m_InitialTime { 0.0 }

private

Settings

Definition at line 170 of file itkStandardStochasticGradientDescentOptimizer.h.

◆ m_Param_A

double itk::StandardStochasticGradientOptimizer::m_Param_A { 1.0 }

private

Definition at line 166 of file itkStandardStochasticGradientDescentOptimizer.h.

◆ m_Param_a

double itk::StandardStochasticGradientOptimizer::m_Param_a { 1.0 }

private

Parameters, as described by Spall.

Definition at line 164 of file itkStandardStochasticGradientDescentOptimizer.h.

◆ m_Param_alpha

double itk::StandardStochasticGradientOptimizer::m_Param_alpha { 0.602 }

private

Definition at line 167 of file itkStandardStochasticGradientDescentOptimizer.h.

◆ m_Param_beta

double itk::StandardStochasticGradientOptimizer::m_Param_beta {}

private

Definition at line 165 of file itkStandardStochasticGradientDescentOptimizer.h.

◆ m_UseConstantStep

bool itk::StandardStochasticGradientOptimizer::m_UseConstantStep {}

protected

Constant step size or others, different value of k.

Definition at line 160 of file itkStandardStochasticGradientDescentOptimizer.h.

Detailed Description

Public Types

Public Member Functions

Static Public Member Functions

Protected Member Functions

Protected Attributes

Private Attributes

Additional Inherited Members

Member Typedef Documentation

◆ ConstPointer

◆ Pointer

◆ Self

◆ Superclass

Member Enumeration Documentation

◆ StopConditionType

Constructor & Destructor Documentation

◆ StandardStochasticGradientOptimizer()

◆ ~StandardStochasticGradientOptimizer()

Member Function Documentation

◆ AdvanceOneStep()

◆ Compute_a()

◆ Compute_beta()

◆ GetCurrentTime()

◆ GetInitialTime()

◆ GetParam_A()

◆ GetParam_a()

◆ GetParam_alpha()

◆ GetParam_beta()

◆ ITK_DISALLOW_COPY_AND_MOVE()

◆ itkOverrideGetNameOfClassMacro()

◆ New()

◆ ResetCurrentTimeToInitialTime()

◆ SetInitialTime()

◆ SetParam_A()

◆ SetParam_a()

◆ SetParam_alpha()

◆ SetParam_beta()

◆ StartOptimization()

◆ UpdateCurrentTime()

Member Data Documentation

◆ m_CurrentTime

◆ m_InitialTime

◆ m_Param_A

◆ m_Param_a

◆ m_Param_alpha

◆ m_Param_beta

◆ m_UseConstantStep