This class does problem scaling by setting the scaling parameters based on the maximum of the gradient at the user provided initial point. More...

#include <IpGradientScaling.hpp>

Inheritance diagram for Ipopt::GradientScaling:

Public Member Functions
Constructors/Destructors
	GradientScaling (const SmartPtr< NLP > &nlp)

virtual	~GradientScaling ()
	Destructor.

Public Member Functions inherited from Ipopt::StandardScalingBase
virtual void	DetermineScaling (const SmartPtr< const VectorSpace > x_space, const SmartPtr< const VectorSpace > c_space, const SmartPtr< const VectorSpace > d_space, const SmartPtr< const MatrixSpace > jac_c_space, const SmartPtr< const MatrixSpace > jac_d_space, const SmartPtr< const SymMatrixSpace > h_space, SmartPtr< const MatrixSpace > &new_jac_c_space, SmartPtr< const MatrixSpace > &new_jac_d_space, SmartPtr< const SymMatrixSpace > &new_h_space, const Matrix &Px_L, const Vector &x_L, const Matrix &Px_U, const Vector &x_U)
	This method is called by the IpoptNLP's at a convenient time to compute and/or read scaling factors.

	StandardScalingBase ()

virtual	~StandardScalingBase ()
	Destructor.

virtual Number	apply_obj_scaling (const Number &f)
	Methods to map scaled and unscaled matrices.

virtual Number	unapply_obj_scaling (const Number &f)
	Returns an obj-unscaled version of the given scalar.

virtual SmartPtr< Vector >	apply_vector_scaling_x_NonConst (const SmartPtr< const Vector > &v)
	Returns an x-scaled version of the given vector.

virtual SmartPtr< const Vector >	apply_vector_scaling_x (const SmartPtr< const Vector > &v)
	Returns an x-scaled version of the given vector.

virtual SmartPtr< Vector >	unapply_vector_scaling_x_NonConst (const SmartPtr< const Vector > &v)
	Returns an x-unscaled version of the given vector.

virtual SmartPtr< const Vector >	unapply_vector_scaling_x (const SmartPtr< const Vector > &v)
	Returns an x-unscaled version of the given vector.

virtual SmartPtr< const Vector >	apply_vector_scaling_c (const SmartPtr< const Vector > &v)
	Returns an c-scaled version of the given vector.

virtual SmartPtr< const Vector >	unapply_vector_scaling_c (const SmartPtr< const Vector > &v)
	Returns an c-unscaled version of the given vector.

virtual SmartPtr< Vector >	apply_vector_scaling_c_NonConst (const SmartPtr< const Vector > &v)
	Returns an c-scaled version of the given vector.

virtual SmartPtr< Vector >	unapply_vector_scaling_c_NonConst (const SmartPtr< const Vector > &v)
	Returns an c-unscaled version of the given vector.

virtual SmartPtr< const Vector >	apply_vector_scaling_d (const SmartPtr< const Vector > &v)
	Returns an d-scaled version of the given vector.

virtual SmartPtr< const Vector >	unapply_vector_scaling_d (const SmartPtr< const Vector > &v)
	Returns an d-unscaled version of the given vector.

virtual SmartPtr< Vector >	apply_vector_scaling_d_NonConst (const SmartPtr< const Vector > &v)
	Returns an d-scaled version of the given vector.

virtual SmartPtr< Vector >	unapply_vector_scaling_d_NonConst (const SmartPtr< const Vector > &v)
	Returns an d-unscaled version of the given vector.

virtual SmartPtr< const Matrix >	apply_jac_c_scaling (SmartPtr< const Matrix > matrix)
	Returns a scaled version of the Jacobian for c.

virtual SmartPtr< const Matrix >	apply_jac_d_scaling (SmartPtr< const Matrix > matrix)
	Returns a scaled version of the Jacobian for d.

virtual SmartPtr< const SymMatrix >	apply_hessian_scaling (SmartPtr< const SymMatrix > matrix)
	Returns a scaled version of the Hessian of the Lagrangian.

virtual bool	have_x_scaling ()
	Returns true if the primal x variables are scaled.

virtual bool	have_c_scaling ()
	Returns true if the equality constraints are scaled.

virtual bool	have_d_scaling ()
	Returns true if the inequality constraints are scaled.

Public Member Functions inherited from Ipopt::NLPScalingObject
bool	Initialize (const Journalist &jnlst, const OptionsList &options, const std::string &prefix)
	Method to initialize the options.

	NLPScalingObject ()

virtual	~NLPScalingObject ()
	Destructor.

SmartPtr< Vector >	apply_vector_scaling_x_LU_NonConst (const Matrix &Px_LU, const SmartPtr< const Vector > &lu, const VectorSpace &x_space)
	Methods for scaling bounds - these wrap those above.

SmartPtr< const Vector >	apply_vector_scaling_x_LU (const Matrix &Px_LU, const SmartPtr< const Vector > &lu, const VectorSpace &x_space)
	Returns an x-scaled vector in the x_L or x_U space.

SmartPtr< Vector >	apply_vector_scaling_d_LU_NonConst (const Matrix &Pd_LU, const SmartPtr< const Vector > &lu, const VectorSpace &d_space)
	Returns an d-scaled vector in the d_L or d_U space.

SmartPtr< const Vector >	apply_vector_scaling_d_LU (const Matrix &Pd_LU, const SmartPtr< const Vector > &lu, const VectorSpace &d_space)
	Returns an d-scaled vector in the d_L or d_U space.

SmartPtr< Vector >	unapply_vector_scaling_d_LU_NonConst (const Matrix &Pd_LU, const SmartPtr< const Vector > &lu, const VectorSpace &d_space)
	Returns an d-unscaled vector in the d_L or d_U space.

SmartPtr< const Vector >	unapply_vector_scaling_d_LU (const Matrix &Pd_LU, const SmartPtr< const Vector > &lu, const VectorSpace &d_space)
	Returns an d-unscaled vector in the d_L or d_U space.

virtual SmartPtr< Vector >	apply_grad_obj_scaling_NonConst (const SmartPtr< const Vector > &v)
	Methods for scaling the gradient of the objective - wraps the virtual methods above.

virtual SmartPtr< const Vector >	apply_grad_obj_scaling (const SmartPtr< const Vector > &v)
	Returns a grad_f scaled version (d_f * D_x^{-1}) of the given vector.

virtual SmartPtr< Vector >	unapply_grad_obj_scaling_NonConst (const SmartPtr< const Vector > &v)
	Returns a grad_f unscaled version (d_f * D_x^{-1}) of the given vector.

virtual SmartPtr< const Vector >	unapply_grad_obj_scaling (const SmartPtr< const Vector > &v)
	Returns a grad_f unscaled version (d_f * D_x^{-1}) of the given vector.

Public Member Functions inherited from Ipopt::ReferencedObject
	ReferencedObject ()

virtual	~ReferencedObject ()

Index	ReferenceCount () const

void	AddRef (const Referencer *referencer) const

void	ReleaseRef (const Referencer *referencer) const

Static Public Member Functions
static void	RegisterOptions (const SmartPtr< RegisteredOptions > &roptions)
	Register the options for this class.

Static Public Member Functions inherited from Ipopt::StandardScalingBase
static void	RegisterOptions (SmartPtr< RegisteredOptions > roptions)

Protected Member Functions
bool	InitializeImpl (const OptionsList &options, const std::string &prefix)
	Initialize the object from the options.

virtual void	DetermineScalingParametersImpl (const SmartPtr< const VectorSpace > x_space, const SmartPtr< const VectorSpace > c_space, const SmartPtr< const VectorSpace > d_space, const SmartPtr< const MatrixSpace > jac_c_space, const SmartPtr< const MatrixSpace > jac_d_space, const SmartPtr< const SymMatrixSpace > h_space, const Matrix &Px_L, const Vector &x_L, const Matrix &Px_U, const Vector &x_U, Number &df, SmartPtr< Vector > &dx, SmartPtr< Vector > &dc, SmartPtr< Vector > &dd)
	This is the method that has to be overloaded by a particular scaling method that somehow computes the scaling vectors dx, dc, and dd.

Protected Member Functions inherited from Ipopt::NLPScalingObject
const Journalist &	Jnlst () const
	Accessor method for the journalist.

Private Member Functions
Default Compiler Generated Methods
(Hidden to avoid implicit creation/calling). These methods are not implemented and we do not want the compiler to implement them for us, so we declare them private and do not define them. This ensures that they will not be implicitly created/called.
	GradientScaling (const GradientScaling &)
	Copy Constructor.

void	operator= (const GradientScaling &)
	Default Assignment Operator.

Private Attributes
SmartPtr< NLP >	nlp_
	pointer to the NLP to get scaling parameters

Number	scaling_max_gradient_
	maximum allowed gradient before scaling is performed

Number	scaling_obj_target_gradient_
	target size of norm for objective gradient

Number	scaling_constr_target_gradient_
	target size of norm for constraint gradients

Number	scaling_min_value_
	minimum value of a scaling parameter

Detailed Description

This class does problem scaling by setting the scaling parameters based on the maximum of the gradient at the user provided initial point.

Definition at line 20 of file IpGradientScaling.hpp.

Constructor & Destructor Documentation

◆ GradientScaling() [1/2]

Ipopt::GradientScaling::GradientScaling ( const SmartPtr< NLP > & nlp )

inline

Definition at line 25 of file IpGradientScaling.hpp.

◆ ~GradientScaling()

virtual Ipopt::GradientScaling::~GradientScaling ( )

inlinevirtual

Destructor.

Definition at line 34 of file IpGradientScaling.hpp.

◆ GradientScaling() [2/2]

Ipopt::GradientScaling::GradientScaling ( const GradientScaling & )

private

Copy Constructor.

Member Function Documentation

◆ RegisterOptions()

static void Ipopt::GradientScaling::RegisterOptions ( const SmartPtr< RegisteredOptions > & roptions )

static

Register the options for this class.

◆ InitializeImpl()

bool Ipopt::GradientScaling::InitializeImpl	(	const OptionsList &	options,
		const std::string &	prefix
	)

protectedvirtual

Initialize the object from the options.

Reimplemented from Ipopt::StandardScalingBase.

◆ DetermineScalingParametersImpl()

virtual void Ipopt::GradientScaling::DetermineScalingParametersImpl	(	const SmartPtr< const VectorSpace >	x_space,
		const SmartPtr< const VectorSpace >	c_space,
		const SmartPtr< const VectorSpace >	d_space,
		const SmartPtr< const MatrixSpace >	jac_c_space,
		const SmartPtr< const MatrixSpace >	jac_d_space,
		const SmartPtr< const SymMatrixSpace >	h_space,
		const Matrix &	Px_L,
		const Vector &	x_L,
		const Matrix &	Px_U,
		const Vector &	x_U,
		Number &	df,
		SmartPtr< Vector > &	dx,
		SmartPtr< Vector > &	dc,
		SmartPtr< Vector > &	dd
	)

protectedvirtual

This is the method that has to be overloaded by a particular scaling method that somehow computes the scaling vectors dx, dc, and dd.

The pointers to those vectors can be NULL, in which case no scaling for that item will be done later.

Implements Ipopt::StandardScalingBase.

◆ operator=()

void Ipopt::GradientScaling::operator= ( const GradientScaling & )

private

Default Assignment Operator.

Member Data Documentation

◆ nlp_

SmartPtr<NLP> Ipopt::GradientScaling::nlp_

private

pointer to the NLP to get scaling parameters

Definition at line 91 of file IpGradientScaling.hpp.

◆ scaling_max_gradient_

Number Ipopt::GradientScaling::scaling_max_gradient_

private

maximum allowed gradient before scaling is performed

Definition at line 94 of file IpGradientScaling.hpp.

◆ scaling_obj_target_gradient_

Number Ipopt::GradientScaling::scaling_obj_target_gradient_

private

target size of norm for objective gradient

Definition at line 97 of file IpGradientScaling.hpp.

◆ scaling_constr_target_gradient_

Number Ipopt::GradientScaling::scaling_constr_target_gradient_

private

target size of norm for constraint gradients

Definition at line 100 of file IpGradientScaling.hpp.

◆ scaling_min_value_

Number Ipopt::GradientScaling::scaling_min_value_

private

minimum value of a scaling parameter

Definition at line 103 of file IpGradientScaling.hpp.

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

src/Algorithm/IpGradientScaling.hpp

Public Member Functions

Static Public Member Functions

Protected Member Functions

Private Member Functions

Private Attributes

Detailed Description

Constructor & Destructor Documentation

◆ GradientScaling() [1/2]

◆ ~GradientScaling()

◆ GradientScaling() [2/2]

Member Function Documentation

◆ RegisterOptions()

◆ InitializeImpl()

◆ DetermineScalingParametersImpl()

◆ operator=()

Member Data Documentation

◆ nlp_

◆ scaling_max_gradient_

◆ scaling_obj_target_gradient_

◆ scaling_constr_target_gradient_

◆ scaling_min_value_