itk::ScaledSingleValuedNonLinearOptimizer Class Reference

#include <itkScaledSingleValuedNonLinearOptimizer.h>

Detailed Description

Optimizers that inherit from this class optimize a scaled cost function instead of the original function :

where are the scaled parameters, the original parameters and the scales.

During optimization the inheriting classes should update the ScaledCurrentPosition ( ) instead of the CurrentPosition ( ).

When an optimizer needs the value at a (scaled) position , it should use the function this->GetScaledValue( ) instead of the GetValue method. Similar for the derivative.

Typically, in StartOptimization() the following line should be present:
this->SetCurrentPosition(this->GetInitialPosition);
This makes sure that the initial position , where is the initial (unscaled) position entered by the user.

Note that:

• GetScaledCurrentPosition returns the current .
• GetCurrentPosition returns the current . This array is only computed when asked for by the user.
• It is NOT necessary to set the CurrentPosition!! In fact, it is not possible anymore: the SetCurrentPosition(param) method is overrided and calls SetScaledCurrentPositon(param*scales).
• The ITK convention is to set the squared scales in the optimizer. So, if you want a scaling s, you must call SetScales( ) (where .* symbolises the element-wise product of with )

Definition at line 62 of file itkScaledSingleValuedNonLinearOptimizer.h.

Inheritance diagram for itk::ScaledSingleValuedNonLinearOptimizer:

Public Types
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
virtual const char *	GetClassName () const
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)
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 ()

Protected Member Functions
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
ScaledCostFunctionPointer	m_ScaledCostFunction {}
ParametersType	m_ScaledCurrentPosition {}

Private Attributes
bool	m_Maximize {}
ParametersType	m_UnscaledCurrentPosition {}

Member Typedef Documentation

ConstPointer

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

Definition at line 71 of file itkScaledSingleValuedNonLinearOptimizer.h.

Pointer

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

Definition at line 70 of file itkScaledSingleValuedNonLinearOptimizer.h.

ScaledCostFunctionPointer

using itk::ScaledSingleValuedNonLinearOptimizer::ScaledCostFunctionPointer = ScaledCostFunctionType::Pointer

Definition at line 87 of file itkScaledSingleValuedNonLinearOptimizer.h.

ScaledCostFunctionType

using itk::ScaledSingleValuedNonLinearOptimizer::ScaledCostFunctionType = ScaledSingleValuedCostFunction

Definition at line 86 of file itkScaledSingleValuedNonLinearOptimizer.h.

ScalesType

using itk::ScaledSingleValuedNonLinearOptimizer::ScalesType = NonLinearOptimizer::ScalesType

Definition at line 85 of file itkScaledSingleValuedNonLinearOptimizer.h.

Self

using itk::ScaledSingleValuedNonLinearOptimizer::Self = ScaledSingleValuedNonLinearOptimizer

Standard ITK-stuff.

Definition at line 68 of file itkScaledSingleValuedNonLinearOptimizer.h.

Superclass

using itk::ScaledSingleValuedNonLinearOptimizer::Superclass = SingleValuedNonLinearOptimizer

Definition at line 69 of file itkScaledSingleValuedNonLinearOptimizer.h.

Constructor & Destructor Documentation

ScaledSingleValuedNonLinearOptimizer()

itk::ScaledSingleValuedNonLinearOptimizer::ScaledSingleValuedNonLinearOptimizer ( )

protected

The constructor.

~ScaledSingleValuedNonLinearOptimizer()

itk::ScaledSingleValuedNonLinearOptimizer::~ScaledSingleValuedNonLinearOptimizer ( )

overrideprotecteddefault

The destructor.

Member Function Documentation

GetClassName()

virtual const char * itk::ScaledSingleValuedNonLinearOptimizer::GetClassName ( ) const

virtual

Run-time type information (and related methods).

Reimplemented in elastix::AdaGrad< TElastix >, elastix::AdaptiveStochasticGradientDescent< TElastix >, elastix::AdaptiveStochasticLBFGS< TElastix >, elastix::AdaptiveStochasticVarianceReducedGradient< TElastix >, elastix::CMAEvolutionStrategy< TElastix >, elastix::ConjugateGradient< TElastix >, elastix::FiniteDifferenceGradientDescent< TElastix >, elastix::PreconditionedStochasticGradientDescent< TElastix >, elastix::QuasiNewtonLBFGS< TElastix >, elastix::StandardGradientDescent< TElastix >, itk::AdaptiveStepsizeOptimizer, itk::AdaptiveStochasticGradientDescentOptimizer, itk::AdaptiveStochasticLBFGSOptimizer, itk::AdaptiveStochasticVarianceReducedGradientOptimizer, itk::CMAEvolutionStrategyOptimizer, itk::ComputeDisplacementDistribution< TFixedImage, TTransform >, itk::ComputePreconditionerUsingDisplacementDistribution< TFixedImage, TTransform >, itk::FiniteDifferenceGradientDescentOptimizer, itk::GenericConjugateGradientOptimizer, itk::GradientDescentOptimizer2, itk::PreconditionedASGDOptimizer, itk::QuasiNewtonLBFGSOptimizer, itk::StandardGradientDescentOptimizer, itk::StandardStochasticGradientOptimizer, itk::StandardStochasticVarianceReducedGradientOptimizer, itk::StochasticGradientDescentOptimizer, and itk::StochasticVarianceReducedGradientDescentOptimizer.

GetCurrentPosition()

const ParametersType & itk::ScaledSingleValuedNonLinearOptimizer::GetCurrentPosition ( ) const

override

Get the current unscaled position: get the ScaledCurrentPosition and divide each element through its scale.

GetMaximize()

virtual bool itk::ScaledSingleValuedNonLinearOptimizer::GetMaximize ( ) const

virtual

GetScaledCostFunction()

virtual const ScaledCostFunctionType * itk::ScaledSingleValuedNonLinearOptimizer::GetScaledCostFunction ( )

virtual

Get a pointer to the scaled cost function.

GetScaledCurrentPosition()

virtual const ParametersType & itk::ScaledSingleValuedNonLinearOptimizer::GetScaledCurrentPosition ( )

virtual

Get the current scaled position.

GetScaledDerivative()

virtual void itk::ScaledSingleValuedNonLinearOptimizer::GetScaledDerivative ( const ParametersType & parameters, DerivativeType & derivative ) const

protectedvirtual

Divide the (scaled) parameters by the scales, call the GetDerivative routine of the unscaled cost function and divide the resulting derivative by the scales.

GetScaledValue()

virtual MeasureType itk::ScaledSingleValuedNonLinearOptimizer::GetScaledValue ( const ParametersType & parameters ) const

protectedvirtual

Divide the (scaled) parameters by the scales and call the GetValue routine of the unscaled cost function.

GetScaledValueAndDerivative()

virtual void itk::ScaledSingleValuedNonLinearOptimizer::GetScaledValueAndDerivative ( const ParametersType & parameters, MeasureType & value, DerivativeType & derivative ) const

protectedvirtual

Same procedure as in GetValue and GetDerivative.

GetUseScales()

bool itk::ScaledSingleValuedNonLinearOptimizer::GetUseScales

(

)

const

InitializeScales()

virtual void itk::ScaledSingleValuedNonLinearOptimizer::InitializeScales ( )

virtual

Configure the scaled cost function. This function sets the current scales in the ScaledCostFunction. NB: it assumes that the scales entered by the user are the squared scales (following the ITK convention). Call this method in StartOptimization() and after entering new scales.

ITK_DISALLOW_COPY_AND_MOVE()

itk::ScaledSingleValuedNonLinearOptimizer::ITK_DISALLOW_COPY_AND_MOVE

(

ScaledSingleValuedNonLinearOptimizer

)

MaximizeOff()

virtual void itk::ScaledSingleValuedNonLinearOptimizer::MaximizeOff ( )

virtual

MaximizeOn()

virtual void itk::ScaledSingleValuedNonLinearOptimizer::MaximizeOn ( )

virtual

Setting: set to 'true' if you want to maximize the cost function. It forces the scaledCostFunction to negate the cost function value and its derivative.

New()

static Pointer itk::ScaledSingleValuedNonLinearOptimizer::New ( )

static

Method for creation through the object factory.

PrintSelf()

void itk::ScaledSingleValuedNonLinearOptimizer::PrintSelf ( std::ostream & os, Indent indent ) const

overrideprotected

PrintSelf.

SetCostFunction()

void itk::ScaledSingleValuedNonLinearOptimizer::SetCostFunction ( CostFunctionType * costFunction )

override

Setting: SetCostFunction.

SetCurrentPosition()

void itk::ScaledSingleValuedNonLinearOptimizer::SetCurrentPosition ( const ParametersType & param )

overrideprotected

Set the scaled current position by entering the non-scaled parameters. The method multiplies param by the scales and calls SetScaledCurrentPosition.

Note: It is not possible (and needed) anymore to set m_CurrentPosition. Optimizers that inherit from this class should optimize the scaled parameters!

This method will probably only be used to convert the InitialPosition entered by the user.

SetMaximize()

virtual void itk::ScaledSingleValuedNonLinearOptimizer::SetMaximize ( bool _arg )

virtual

SetScaledCurrentPosition()

virtual void itk::ScaledSingleValuedNonLinearOptimizer::SetScaledCurrentPosition ( const ParametersType & parameters )

protectedvirtual

Set m_ScaledCurrentPosition.

SetUseScales()

virtual void itk::ScaledSingleValuedNonLinearOptimizer::SetUseScales ( bool arg )

virtual

Setting: Turn on/off the use of scales. Set this flag to false when no scaling is desired.

Field Documentation

m_Maximize

bool itk::ScaledSingleValuedNonLinearOptimizer::m_Maximize {}

private

Definition at line 191 of file itkScaledSingleValuedNonLinearOptimizer.h.

m_ScaledCostFunction

ScaledCostFunctionPointer itk::ScaledSingleValuedNonLinearOptimizer::m_ScaledCostFunction {}

protected

Definition at line 146 of file itkScaledSingleValuedNonLinearOptimizer.h.

m_ScaledCurrentPosition

ParametersType itk::ScaledSingleValuedNonLinearOptimizer::m_ScaledCurrentPosition {}

protected

Member variables.

Definition at line 145 of file itkScaledSingleValuedNonLinearOptimizer.h.

m_UnscaledCurrentPosition

ParametersType itk::ScaledSingleValuedNonLinearOptimizer::m_UnscaledCurrentPosition {}

mutableprivate

Variable to store the CurrentPosition, when the function GetCurrentPosition is called. This method needs a member variable, because the GetCurrentPosition return something by reference.

Definition at line 190 of file itkScaledSingleValuedNonLinearOptimizer.h.