Class for handling the perturbation factors delta_x, delta_s, delta_c, and delta_d in the primal dual system. More...

#include <IpCGPerturbationHandler.hpp>

Inheritance diagram for Ipopt::CGPerturbationHandler:

Public Member Functions
virtual bool	InitializeImpl (const OptionsList &options, const std::string &prefix)
	Implementation of the initialization method that has to be overloaded by for each derived class. More...

bool	ConsiderNewSystem (Number &delta_x, Number &delta_s, Number &delta_c, Number &delta_d)
	This method must be called for each new matrix, and before any other method for generating perturbation factors. More...

bool	PerturbForSingularity (Number &delta_x, Number &delta_s, Number &delta_c, Number &delta_d)
	This method returns pertubation factors for the case when the most recent factorization resulted in a singular matrix. More...

bool	PerturbForWrongInertia (Number &delta_x, Number &delta_s, Number &delta_c, Number &delta_d)
	This method returns pertubation factors for the case when the most recent factorization resulted in a matrix with an incorrect number of negative eigenvalues. More...

void	CurrentPerturbation (Number &delta_x, Number &delta_s, Number &delta_c, Number &delta_d)
	Just return the perturbation values that have been determined most recently. More...

Constructors/Destructors
	CGPerturbationHandler ()
	Default Constructor. More...

virtual	~CGPerturbationHandler ()
	Default destructor. More...

Public Member Functions inherited from Ipopt::PDPerturbationHandler
	PDPerturbationHandler ()
	Default Constructor. More...

virtual	~PDPerturbationHandler ()
	Default destructor. More...

Public Member Functions inherited from Ipopt::AlgorithmStrategyObject
bool	Initialize (const Journalist &jnlst, IpoptNLP &ip_nlp, IpoptData &ip_data, IpoptCalculatedQuantities &ip_cq, const OptionsList &options, const std::string &prefix)
	This method is called every time the algorithm starts again - it is used to reset any internal state. More...

bool	ReducedInitialize (const Journalist &jnlst, const OptionsList &options, const std::string &prefix)
	Reduced version of the Initialize method, which does not require special Ipopt information. More...

	AlgorithmStrategyObject ()
	Default Constructor. More...

virtual	~AlgorithmStrategyObject ()
	Default Destructor. More...

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 (SmartPtr< RegisteredOptions > roptions)
	Methods for IpoptType. More...

Static Public Member Functions inherited from Ipopt::PDPerturbationHandler
static void	RegisterOptions (SmartPtr< RegisteredOptions > roptions)
	Methods for IpoptType. More...

Private Member Functions
CGPenaltyData &	CGPenData ()
	Method to easily access CGPenalty data. More...

CGPenaltyCq &	CGPenCq ()
	Method to easily access CGPenalty calculated quantities. More...

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.
	CGPerturbationHandler (const CGPerturbationHandler &)
	Copy Constructor. More...

void	operator= (const CGPerturbationHandler &)
	Overloaded Equals Operator. More...

Auxilliary methods
bool	get_deltas_for_wrong_inertia (Number &delta_x, Number &delta_s, Number &delta_c, Number &delta_d)
	Internal version of PerturbForWrongInertia with the difference, that finalize_test is not called. More...

void	finalize_test ()
	This method is call whenever a matrix had been factorization and is not singular. More...

Number	delta_cd ()
	Compute perturbation value for constraints. More...

Private Attributes
bool	get_deltas_for_wrong_inertia_called_
	Flag indicating if for the given matrix the perturb for wrong inertia method has already been called. More...

Number	penalty_max_
	The max reference value for scaling the penalty parameter. More...

Number	mult_diverg_feasibility_tol_
	Feasibility for perturbation in pure Newton method. More...

Size of the most recent non-zero perturbation.
Number	delta_x_last_
	The last nonzero value for delta_x. More...

Number	delta_s_last_
	The last nonzero value for delta_s. More...

Number	delta_c_last_
	The last nonzero value for delta_c. More...

Number	delta_d_last_
	The last nonzero value for delta_d. More...

Size of the most recently suggested perturbation for the
current matrix.
Number	delta_x_curr_
	The current value for delta_x. More...

Number	delta_s_curr_
	The current value for delta_s. More...

Number	delta_c_curr_
	The current value for delta_c. More...

Number	delta_d_curr_
	The current value for delta_d. More...

Algorithmic parameters.
Number	delta_xs_max_
	Maximal perturbation for x and s. More...

Number	delta_xs_min_
	Smallest possible perturbation for x and s. More...

Number	delta_xs_first_inc_fact_
	Increase factor for delta_xs for first required perturbation. More...

Number	delta_xs_inc_fact_
	Increase factor for delta_xs for later perturbations. More...

Number	delta_xs_dec_fact_
	Decrease factor for delta_xs for later perturbations. More...

Number	delta_xs_init_
	Very first trial value for delta_xs perturbation. More...

Number	delta_cd_val_
	Size of perturbation for c and d blocks. More...

Number	delta_cd_exp_
	Exponent on mu in formula for of perturbation for c and d blocks. More...

bool	reset_last_
	Flag indicating whether the new values are based on the perturbations in the last iteration or in the more recent iteration in which a perturbation was done. More...

Index	degen_iters_max_
	Required number of iterations for degeneracy conclusions. More...

bool	perturb_always_cd_
	Flag indicating that the delta_c, delta_d perturbation should always be used. More...

Handling structural degeneracy
enum	DegenType { NOT_YET_DETERMINED, NOT_DEGENERATE, DEGENERATE }
	Type for degeneracy flags. More...

enum	TrialStatus { NO_TEST, TEST_DELTA_C_EQ_0_DELTA_X_EQ_0, TEST_DELTA_C_GT_0_DELTA_X_EQ_0, TEST_DELTA_C_EQ_0_DELTA_X_GT_0, TEST_DELTA_C_GT_0_DELTA_X_GT_0 }
	Status of current trial configuration. More...

DegenType	hess_degenerate_
	Flag indicating whether the reduced Hessian matrix is thought to be structurally singular. More...

DegenType	jac_degenerate_
	Flag indicating whether the Jacobian of the constraints is thought to be structurally rank-deficient. More...

Index	degen_iters_
	Flag counting matrices in which degeneracy was observed in the first successive iterations. More...

TrialStatus	test_status_
	Current status. More...

Additional Inherited Members
Protected Types inherited from Ipopt::PDPerturbationHandler
enum	DegenType { NOT_YET_DETERMINED, NOT_DEGENERATE, DEGENERATE }
	Type for degeneracy flags. More...

enum	TrialStatus { NO_TEST, TEST_DELTA_C_EQ_0_DELTA_X_EQ_0, TEST_DELTA_C_GT_0_DELTA_X_EQ_0, TEST_DELTA_C_EQ_0_DELTA_X_GT_0, TEST_DELTA_C_GT_0_DELTA_X_GT_0 }
	Status of current trial configuration. More...

Protected Member Functions inherited from Ipopt::PDPerturbationHandler
	PDPerturbationHandler (const PDPerturbationHandler &)
	Copy Constructor. More...

void	operator= (const PDPerturbationHandler &)
	Overloaded Equals Operator. More...

bool	get_deltas_for_wrong_inertia (Number &delta_x, Number &delta_s, Number &delta_c, Number &delta_d)
	Internal version of PerturbForWrongInertia with the difference, that finalize_test is not called. More...

void	finalize_test ()
	This method is call whenever a matrix had been factorization and is not singular. More...

Number	delta_cd ()
	Compute perturbation value for constraints. More...

Protected Member Functions inherited from Ipopt::AlgorithmStrategyObject
const Journalist &	Jnlst () const

IpoptNLP &	IpNLP () const

IpoptData &	IpData () const

IpoptCalculatedQuantities &	IpCq () const

bool	HaveIpData () const

Protected Attributes inherited from Ipopt::PDPerturbationHandler
bool	get_deltas_for_wrong_inertia_called_
	Flag indicating if for the given matrix the perturb for wrong inertia method has already been called. More...

Number	delta_x_last_
	The last nonzero value for delta_x. More...

Number	delta_s_last_
	The last nonzero value for delta_s. More...

Number	delta_c_last_
	The last nonzero value for delta_c. More...

Number	delta_d_last_
	The last nonzero value for delta_d. More...

Number	delta_x_curr_
	The current value for delta_x. More...

Number	delta_s_curr_
	The current value for delta_s. More...

Number	delta_c_curr_
	The current value for delta_c. More...

Number	delta_d_curr_
	The current value for delta_d. More...

Number	delta_xs_max_
	Maximal perturbation for x and s. More...

Number	delta_xs_min_
	Smallest possible perturbation for x and s. More...

Number	delta_xs_first_inc_fact_
	Increase factor for delta_xs for first required perturbation. More...

Number	delta_xs_inc_fact_
	Increase factor for delta_xs for later perturbations. More...

Number	delta_xs_dec_fact_
	Decrease factor for delta_xs for later perturbations. More...

Number	delta_xs_init_
	Very first trial value for delta_xs perturbation. More...

Number	delta_cd_val_
	Size of perturbation for c and d blocks. More...

Number	delta_cd_exp_
	Exponent on mu in formula for of perturbation for c and d blocks. More...

bool	reset_last_
	Flag indicating whether the new values are based on the perturbations in the last iteration or in the more recent iteration in which a perturbation was done. More...

Index	degen_iters_max_
	Required number of iterations for degeneracy conclusions. More...

bool	perturb_always_cd_
	Flag indicating that the delta_c, delta_d perturbation should always be used. More...

DegenType	hess_degenerate_
	Flag indicating whether the reduced Hessian matrix is thought to be structurally singular. More...

DegenType	jac_degenerate_
	Flag indicating whether the Jacobian of the constraints is thought to be structurally rank-deficient. More...

Index	degen_iters_
	Flag counting matrices in which degeneracy was observed in the first successive iterations. More...

TrialStatus	test_status_
	Current status. More...

Detailed Description

Class for handling the perturbation factors delta_x, delta_s, delta_c, and delta_d in the primal dual system.

This class is used by the PDFullSpaceSolver to handle the cases where the primal-dual system is singular or has the wrong inertia. The perturbation factors are obtained based on simple heuristics, taking into account the size of previous perturbations.

Definition at line 25 of file IpCGPerturbationHandler.hpp.

Member Enumeration Documentation

enum Ipopt::CGPerturbationHandler::DegenType

private

Type for degeneracy flags.

Enumerator
NOT_YET_DETERMINED
NOT_DEGENERATE
DEGENERATE

Definition at line 140 of file IpCGPerturbationHandler.hpp.

enum Ipopt::CGPerturbationHandler::TrialStatus

private

Status of current trial configuration.

Enumerator
NO_TEST
TEST_DELTA_C_EQ_0_DELTA_X_EQ_0
TEST_DELTA_C_GT_0_DELTA_X_EQ_0
TEST_DELTA_C_EQ_0_DELTA_X_GT_0
TEST_DELTA_C_GT_0_DELTA_X_GT_0

Definition at line 161 of file IpCGPerturbationHandler.hpp.

Constructor & Destructor Documentation

Ipopt::CGPerturbationHandler::CGPerturbationHandler ( )

Default Constructor.

virtual Ipopt::CGPerturbationHandler::~CGPerturbationHandler ( )

inlinevirtual

Default destructor.

Definition at line 33 of file IpCGPerturbationHandler.hpp.

Ipopt::CGPerturbationHandler::CGPerturbationHandler ( const CGPerturbationHandler & )

private

Copy Constructor.

Member Function Documentation

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

virtual

Implementation of the initialization method that has to be overloaded by for each derived class.

Reimplemented from Ipopt::PDPerturbationHandler.

bool Ipopt::CGPerturbationHandler::ConsiderNewSystem	(	Number &	delta_x,
		Number &	delta_s,
		Number &	delta_c,
		Number &	delta_d
	)

virtual

This method must be called for each new matrix, and before any other method for generating perturbation factors.

Usually, the returned perturbation factors are zero, but if the system is thought to be structurally singular, they might be positive. If the return value is false, no suitable perturbation could be found.

Reimplemented from Ipopt::PDPerturbationHandler.

bool Ipopt::CGPerturbationHandler::PerturbForSingularity	(	Number &	delta_x,
		Number &	delta_s,
		Number &	delta_c,
		Number &	delta_d
	)

virtual

This method returns pertubation factors for the case when the most recent factorization resulted in a singular matrix.

If the return value is false, no suitable perturbation could be found.

Reimplemented from Ipopt::PDPerturbationHandler.

bool Ipopt::CGPerturbationHandler::PerturbForWrongInertia	(	Number &	delta_x,
		Number &	delta_s,
		Number &	delta_c,
		Number &	delta_d
	)

virtual

This method returns pertubation factors for the case when the most recent factorization resulted in a matrix with an incorrect number of negative eigenvalues.

If the return value is false, no suitable perturbation could be found.

Reimplemented from Ipopt::PDPerturbationHandler.

void Ipopt::CGPerturbationHandler::CurrentPerturbation	(	Number &	delta_x,
		Number &	delta_s,
		Number &	delta_c,
		Number &	delta_d
	)

virtual

Just return the perturbation values that have been determined most recently.

Reimplemented from Ipopt::PDPerturbationHandler.

static void Ipopt::CGPerturbationHandler::RegisterOptions ( SmartPtr< RegisteredOptions > roptions )

static

Methods for IpoptType.

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

private

Overloaded Equals Operator.

CGPenaltyData& Ipopt::CGPerturbationHandler::CGPenData ( )

inlineprivate

Method to easily access CGPenalty data.

Definition at line 91 of file IpCGPerturbationHandler.hpp.

CGPenaltyCq& Ipopt::CGPerturbationHandler::CGPenCq ( )

inlineprivate

Method to easily access CGPenalty calculated quantities.

Definition at line 100 of file IpCGPerturbationHandler.hpp.

bool Ipopt::CGPerturbationHandler::get_deltas_for_wrong_inertia	(	Number &	delta_x,
		Number &	delta_s,
		Number &	delta_c,
		Number &	delta_d
	)

private

Internal version of PerturbForWrongInertia with the difference, that finalize_test is not called.

Returns false if the delta_x and delta_s parameters become too large.

void Ipopt::CGPerturbationHandler::finalize_test ( )

private

This method is call whenever a matrix had been factorization and is not singular.

In here, we can evaluate the outcome of the deneracy test heuristics.

Number Ipopt::CGPerturbationHandler::delta_cd ( )

private

Compute perturbation value for constraints.

Member Data Documentation

Number Ipopt::CGPerturbationHandler::delta_x_last_

private

The last nonzero value for delta_x.

Definition at line 111 of file IpCGPerturbationHandler.hpp.

Number Ipopt::CGPerturbationHandler::delta_s_last_

private

The last nonzero value for delta_s.

Definition at line 113 of file IpCGPerturbationHandler.hpp.

Number Ipopt::CGPerturbationHandler::delta_c_last_

private

The last nonzero value for delta_c.

Definition at line 115 of file IpCGPerturbationHandler.hpp.

Number Ipopt::CGPerturbationHandler::delta_d_last_

private

The last nonzero value for delta_d.

Definition at line 117 of file IpCGPerturbationHandler.hpp.

Number Ipopt::CGPerturbationHandler::delta_x_curr_

private

The current value for delta_x.

Definition at line 124 of file IpCGPerturbationHandler.hpp.

Number Ipopt::CGPerturbationHandler::delta_s_curr_

private

The current value for delta_s.

Definition at line 126 of file IpCGPerturbationHandler.hpp.

Number Ipopt::CGPerturbationHandler::delta_c_curr_

private

The current value for delta_c.

Definition at line 128 of file IpCGPerturbationHandler.hpp.

Number Ipopt::CGPerturbationHandler::delta_d_curr_

private

The current value for delta_d.

Definition at line 130 of file IpCGPerturbationHandler.hpp.

bool Ipopt::CGPerturbationHandler::get_deltas_for_wrong_inertia_called_

private

Flag indicating if for the given matrix the perturb for wrong inertia method has already been called.

Definition at line 135 of file IpCGPerturbationHandler.hpp.

DegenType Ipopt::CGPerturbationHandler::hess_degenerate_

private

Flag indicating whether the reduced Hessian matrix is thought to be structurally singular.

Definition at line 149 of file IpCGPerturbationHandler.hpp.

DegenType Ipopt::CGPerturbationHandler::jac_degenerate_

private

Flag indicating whether the Jacobian of the constraints is thought to be structurally rank-deficient.

Definition at line 153 of file IpCGPerturbationHandler.hpp.

Index Ipopt::CGPerturbationHandler::degen_iters_

private

Flag counting matrices in which degeneracy was observed in the first successive iterations.

-1 means that there was a non-degenerate (unperturbed) matrix at some point.

Definition at line 158 of file IpCGPerturbationHandler.hpp.

TrialStatus Ipopt::CGPerturbationHandler::test_status_

private

Current status.

Definition at line 171 of file IpCGPerturbationHandler.hpp.

Number Ipopt::CGPerturbationHandler::delta_xs_max_

private

Maximal perturbation for x and s.

Definition at line 177 of file IpCGPerturbationHandler.hpp.

Number Ipopt::CGPerturbationHandler::delta_xs_min_

private

Smallest possible perturbation for x and s.

Definition at line 179 of file IpCGPerturbationHandler.hpp.

Number Ipopt::CGPerturbationHandler::delta_xs_first_inc_fact_

private

Increase factor for delta_xs for first required perturbation.

Definition at line 181 of file IpCGPerturbationHandler.hpp.

Number Ipopt::CGPerturbationHandler::delta_xs_inc_fact_

private

Increase factor for delta_xs for later perturbations.

Definition at line 183 of file IpCGPerturbationHandler.hpp.

Number Ipopt::CGPerturbationHandler::delta_xs_dec_fact_

private

Decrease factor for delta_xs for later perturbations.

Definition at line 185 of file IpCGPerturbationHandler.hpp.

Number Ipopt::CGPerturbationHandler::delta_xs_init_

private

Very first trial value for delta_xs perturbation.

Definition at line 187 of file IpCGPerturbationHandler.hpp.

Number Ipopt::CGPerturbationHandler::delta_cd_val_

private

Size of perturbation for c and d blocks.

Definition at line 189 of file IpCGPerturbationHandler.hpp.

Number Ipopt::CGPerturbationHandler::delta_cd_exp_

private

Exponent on mu in formula for of perturbation for c and d blocks.

Definition at line 191 of file IpCGPerturbationHandler.hpp.

bool Ipopt::CGPerturbationHandler::reset_last_

private

Flag indicating whether the new values are based on the perturbations in the last iteration or in the more recent iteration in which a perturbation was done.

Definition at line 195 of file IpCGPerturbationHandler.hpp.

Index Ipopt::CGPerturbationHandler::degen_iters_max_

private

Required number of iterations for degeneracy conclusions.

Definition at line 197 of file IpCGPerturbationHandler.hpp.

bool Ipopt::CGPerturbationHandler::perturb_always_cd_

private

Flag indicating that the delta_c, delta_d perturbation should always be used.

Definition at line 200 of file IpCGPerturbationHandler.hpp.

Number Ipopt::CGPerturbationHandler::penalty_max_

private

The max reference value for scaling the penalty parameter.