Ipopt::DefaultIterateInitializer Class Reference

Class implementing the default initialization procedure (based on user options) for the iterates. More...

#include <IpDefaultIterateInitializer.hpp>

Inheritance diagram for Ipopt::DefaultIterateInitializer:

[legend]

Collaboration diagram for Ipopt::DefaultIterateInitializer:

[legend]

List of all members.

Public Types
Enums of option values

enum	BoundMultInitMethod { B_CONSTANT = 0, B_MU_BASED }
Public Member Functions
virtual bool	InitializeImpl (const OptionsList &options, const std::string &prefix)
	overloaded from AlgorithmStrategyObject
virtual bool	SetInitialIterates ()
	Compute the initial iterates and set the into the curr field of the ip_data object.
Constructors/Destructors

	DefaultIterateInitializer (const SmartPtr< EqMultiplierCalculator > &eq_mult_calculator, const SmartPtr< IterateInitializer > &warm_start_initializer, const SmartPtr< AugSystemSolver > aug_system_solver=NULL)
	Constructor.
virtual	~DefaultIterateInitializer ()
	Default destructor.
Static Public Member Functions
static void	push_variables (const Journalist &jnlst, Number bound_push, Number bound_frac, std::string name, const Vector &orig_x, SmartPtr< const Vector > &new_x, const Vector &x_L, const Vector &x_U, const Matrix &Px_L, const Matrix &Px_U)
	Auxilliary function for moving the initial point.
static void	least_square_mults (const Journalist &jnlst, IpoptNLP &ip_nlp, IpoptData &ip_data, IpoptCalculatedQuantities &ip_cq, const SmartPtr< EqMultiplierCalculator > &eq_mult_calculator, Number constr_mult_init_max)
	Auxilliary function for computing least_square multipliers.


static void	RegisterOptions (SmartPtr< RegisteredOptions > reg_options)
	Methods for IpoptType.
Private Member Functions
bool	CalculateLeastSquarePrimals (Vector &x_ls, Vector &s_ls)
	Auxilliary method for computing least square primal variables.
bool	CalculateLeastSquareDuals (Vector &zL_new, Vector &zU_new, Vector &vL_new, Vector &vU_new, Vector &yc_new, Vector &yd_new)
	Auxilliary method for computing least square dual variables.
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.
	DefaultIterateInitializer ()
	Default Constructor.
	DefaultIterateInitializer (const DefaultIterateInitializer &)
	Copy Constructor.
void	operator= (const DefaultIterateInitializer &)
	Overloaded Equals Operator.
Private Attributes
SmartPtr< EqMultiplierCalculator >	eq_mult_calculator_
	object to be used for the initialization of the equality constraint multipliers.
SmartPtr< IterateInitializer >	warm_start_initializer_
	object to be used for a warm start initialization
SmartPtr< AugSystemSolver >	aug_system_solver_
	Object for solving the augmented system.
Algorithmic Parameters

Number	bound_push_
	Absolute parameter for bumping x0.
Number	bound_frac_
	Relative parameter for bumping x0.
Number	slack_bound_push_
	Absolute parameter for bumping s0.
Number	slack_bound_frac_
	Relative parameter for bumping s0.
Number	constr_mult_init_max_
	If max-norm of the initial equality constraint multiplier estimate is larger than this, the initial y_* variables are set to zero.
Number	bound_mult_init_val_
	Initial value for all bound mulitpliers.
bool	warm_start_init_point_
	Flag indicating whether warm_start_initializer should be used instead of the default initialization.
bool	least_square_init_primal_
	Flag indicating whether the primal variables should be initialized as least square fit for the linearized constraints.
bool	least_square_init_duals_
	Flag indicating whether all dual variables should be initialized as least square fit for the linearized dual infeasibility.
BoundMultInitMethod	bound_mult_init_method_
	Flag indicating how bound multipliers are initialized.
Number	mu_init_
	Initial value of barrier parameter.

Detailed Description

Class implementing the default initialization procedure (based on user options) for the iterates.

It is used at the very beginning of the optimization for determine the starting point for all variables.

Definition at line 24 of file IpDefaultIterateInitializer.hpp.

Member Enumeration Documentation

enum Ipopt::DefaultIterateInitializer::BoundMultInitMethod

Enumerator:

B_CONSTANT
B_MU_BASED

Definition at line 88 of file IpDefaultIterateInitializer.hpp.

Constructor & Destructor Documentation

Ipopt::DefaultIterateInitializer::DefaultIterateInitializer	(	const SmartPtr< EqMultiplierCalculator > &	eq_mult_calculator,
		const SmartPtr< IterateInitializer > &	warm_start_initializer,
		const SmartPtr< AugSystemSolver >	aug_system_solver = `NULL`
	)

Constructor.

If eq_mult_calculator is not NULL, it will be used to compute the initial values for equality constraint multipliers. If warm_start_initializer is not NULL, it will be used to compute the initial values if the option warm_start_init_point is chosen.

virtual Ipopt::DefaultIterateInitializer::~DefaultIterateInitializer ( ) [inline, virtual]

Default destructor.

Definition at line 40 of file IpDefaultIterateInitializer.hpp.

Ipopt::DefaultIterateInitializer::DefaultIterateInitializer ( ) [private]

Default Constructor.

Ipopt::DefaultIterateInitializer::DefaultIterateInitializer ( const DefaultIterateInitializer & ) [private]

Copy Constructor.

Member Function Documentation

virtual bool Ipopt::DefaultIterateInitializer::InitializeImpl	(	const OptionsList &	options,
		const std::string &	prefix
	)			`[virtual]`

overloaded from AlgorithmStrategyObject

Implements Ipopt::IterateInitializer.

virtual bool Ipopt::DefaultIterateInitializer::SetInitialIterates ( ) [virtual]

Compute the initial iterates and set the into the curr field of the ip_data object.

Implements Ipopt::IterateInitializer.

static void Ipopt::DefaultIterateInitializer::push_variables	(	const Journalist &	jnlst,
		Number	bound_push,
		Number	bound_frac,
		std::string	name,
		const Vector &	orig_x,
		SmartPtr< const Vector > &	new_x,
		const Vector &	x_L,
		const Vector &	x_U,
		const Matrix &	Px_L,
		const Matrix &	Px_U
	)			`[static]`

Auxilliary function for moving the initial point.

This is declared static so that it can also be used from WarmStartIterateInitializer.

static void Ipopt::DefaultIterateInitializer::least_square_mults	(	const Journalist &	jnlst,
		IpoptNLP &	ip_nlp,
		IpoptData &	ip_data,
		IpoptCalculatedQuantities &	ip_cq,
		const SmartPtr< EqMultiplierCalculator > &	eq_mult_calculator,
		Number	constr_mult_init_max
	)			`[static]`

Auxilliary function for computing least_square multipliers.

The multipliers are computed based on the values in the trial fields (current is overwritten). On return, the multipliers are in the trial fields as well. The value of constr_mult_init_max determines if the computed least square estimate should be used, or if the initial multipliers are set to zero.

static void Ipopt::DefaultIterateInitializer::RegisterOptions ( SmartPtr< RegisteredOptions > reg_options ) [static]

Methods for IpoptType.

void Ipopt::DefaultIterateInitializer::operator= ( const DefaultIterateInitializer & ) [private]

Overloaded Equals Operator.

Reimplemented from Ipopt::IterateInitializer.

bool Ipopt::DefaultIterateInitializer::CalculateLeastSquarePrimals	(	Vector &	x_ls,
		Vector &	s_ls
	)			`[private]`

Auxilliary method for computing least square primal variables.

bool Ipopt::DefaultIterateInitializer::CalculateLeastSquareDuals	(	Vector &	zL_new,
		Vector &	zU_new,
		Vector &	vL_new,
		Vector &	vU_new,
		Vector &	yc_new,
		Vector &	yd_new
	)			`[private]`