Ipopt::IpoptAlgorithm Class Reference

The main ipopt algorithm class. More...

#include <IpIpoptAlg.hpp>

Inheritance diagram for Ipopt::IpoptAlgorithm:

[legend]

Collaboration diagram for Ipopt::IpoptAlgorithm:

[legend]

List of all members.

Public Member Functions
virtual bool	InitializeImpl (const OptionsList &options, const std::string &prefix)
	overloaded from AlgorithmStrategyObject
SolverReturn	Optimize ()
	Main solve method.
Constructors/Destructors

	IpoptAlgorithm (const SmartPtr< SearchDirectionCalculator > &search_dir_calculator, const SmartPtr< LineSearch > &line_search, const SmartPtr< MuUpdate > &mu_update, const SmartPtr< ConvergenceCheck > &conv_check, const SmartPtr< IterateInitializer > &iterate_initializer, const SmartPtr< IterationOutput > &iter_output, const SmartPtr< HessianUpdater > &hessian_updater, const SmartPtr< EqMultiplierCalculator > &eq_multiplier_calculator=NULL)
	Constructor.
virtual	~IpoptAlgorithm ()
	Default destructor.
Static Public Member Functions


static void	RegisterOptions (SmartPtr< RegisteredOptions > roptions)
	Methods for IpoptType.
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.
	IpoptAlgorithm ()
	Default Constructor.
	IpoptAlgorithm (const IpoptAlgorithm &)
	Copy Constructor.
void	operator= (const IpoptAlgorithm &)
	Overloaded Equals Operator.
Main steps of the algorthim

void	UpdateHessian ()
	Method for updating the current Hessian.
bool	UpdateBarrierParameter ()
	Method to update the barrier parameter.
bool	ComputeSearchDirection ()
	Method to setup the call to the PDSystemSolver.
void	ComputeAcceptableTrialPoint ()
	Method computing the new iterate (usually vialine search).
void	AcceptTrialPoint ()
	Method for accepting the trial point as the new iteration, possibly after adjusting the variable bounds in the NLP.
void	OutputIteration ()
	Do all the output for one iteration.
void	InitializeIterates ()
	Sets up initial values for the iterates, Corrects the initial values for x and s (force in bounds).
void	PrintProblemStatistics ()
	Print the problem size statistics.
void	ComputeFeasibilityMultipliers ()
	Compute the Lagrangian multipliers for a feasibility problem.
auxilliary functions

void	calc_number_of_bounds (const Vector &x, const Vector &x_L, const Vector &x_U, const Matrix &Px_L, const Matrix &Px_U, Index &n_tot, Index &n_only_lower, Index &n_both, Index &n_only_upper)
	Method for ensuring that the trial multipliers are not too far from the primal estime.
Number	correct_bound_multiplier (const Vector &trial_z, const Vector &trial_slack, const Vector &trial_compl, SmartPtr< const Vector > &new_trial_z)
	Method for ensuring that the trial multipliers are not too far from the primal estime.
Private Attributes
Strategy objects

SmartPtr < SearchDirectionCalculator >	search_dir_calculator_
	The multipler calculator (for y_c and y_d) has to be set only if option recalc_y is set to true.
SmartPtr< LineSearch >	line_search_
	The multipler calculator (for y_c and y_d) has to be set only if option recalc_y is set to true.
SmartPtr< MuUpdate >	mu_update_
	The multipler calculator (for y_c and y_d) has to be set only if option recalc_y is set to true.
SmartPtr< ConvergenceCheck >	conv_check_
	The multipler calculator (for y_c and y_d) has to be set only if option recalc_y is set to true.
SmartPtr< IterateInitializer >	iterate_initializer_
	The multipler calculator (for y_c and y_d) has to be set only if option recalc_y is set to true.
SmartPtr< IterationOutput >	iter_output_
	The multipler calculator (for y_c and y_d) has to be set only if option recalc_y is set to true.
SmartPtr< HessianUpdater >	hessian_updater_
	The multipler calculator (for y_c and y_d) has to be set only if option recalc_y is set to true.
SmartPtr< EqMultiplierCalculator >	eq_multiplier_calculator_
	The multipler calculator (for y_c and y_d) has to be set only if option recalc_y is set to true.
internal flags

bool	skip_print_problem_stats_
	Flag indicating if the statistic should not be printed.
Algorithmic parameters

Number	kappa_sigma_
	safeguard factor for bound multipliers.
bool	recalc_y_
	Flag indicating whether the y multipliers should be recalculated with the eq_mutliplier_calculator object for each new point.
Number	recalc_y_feas_tol_
	Feasibility threshold for recalc_y.
bool	mehrotra_algorithm_
	Flag indicating if we want to do Mehrotras's algorithm.

Detailed Description

The main ipopt algorithm class.

Main Ipopt algorithm class, contains the main optimize method, handles the execution of the optimization. The constructor initializes the data structures through the nlp, and the Optimize method then assumes that everything is initialized and ready to go. After an optimization is complete, the user can access the solution through the passed in ip_data structure. Multiple calls to the Optimize method are allowed as long as the structure of the problem remains the same (i.e. starting point or nlp parameter changes only).

Definition at line 45 of file IpIpoptAlg.hpp.

Constructor & Destructor Documentation

Ipopt::IpoptAlgorithm::IpoptAlgorithm	(	const SmartPtr< SearchDirectionCalculator > &	search_dir_calculator,
		const SmartPtr< LineSearch > &	line_search,
		const SmartPtr< MuUpdate > &	mu_update,
		const SmartPtr< ConvergenceCheck > &	conv_check,
		const SmartPtr< IterateInitializer > &	iterate_initializer,
		const SmartPtr< IterationOutput > &	iter_output,
		const SmartPtr< HessianUpdater > &	hessian_updater,
		const SmartPtr< EqMultiplierCalculator > &	eq_multiplier_calculator = `NULL`
	)

Constructor.

(The IpoptAlgorithm uses smart pointers for these passed-in pieces to make sure that a user of IpoptAlgoroithm cannot pass in an object created on the stack!)

virtual Ipopt::IpoptAlgorithm::~IpoptAlgorithm ( ) [virtual]

Default destructor.

Ipopt::IpoptAlgorithm::IpoptAlgorithm ( ) [private]

Default Constructor.

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

Copy Constructor.

Member Function Documentation

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

overloaded from AlgorithmStrategyObject

Implements Ipopt::AlgorithmStrategyObject.

SolverReturn Ipopt::IpoptAlgorithm::Optimize ( )

Main solve method.

static void Ipopt::IpoptAlgorithm::RegisterOptions ( SmartPtr< RegisteredOptions > roptions ) [static]

Methods for IpoptType.

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

Overloaded Equals Operator.

Reimplemented from Ipopt::AlgorithmStrategyObject.

void Ipopt::IpoptAlgorithm::UpdateHessian ( ) [private]

Method for updating the current Hessian.

This can either just evaluate the exact Hessian (based on the current iterate), or perform a quasi-Newton update.

bool Ipopt::IpoptAlgorithm::UpdateBarrierParameter ( ) [private]

Method to update the barrier parameter.

Returns false, if the algorithm can't continue with the regular procedure and needs to revert to a fallback mechanism in the line search (such as restoration phase)

bool Ipopt::IpoptAlgorithm::ComputeSearchDirection ( ) [private]

Method to setup the call to the PDSystemSolver.

Returns false, if the algorithm can't continue with the regular procedure and needs to revert to a fallback mechanism in the line search (such as restoration phase)

void Ipopt::IpoptAlgorithm::ComputeAcceptableTrialPoint ( ) [private]

Method computing the new iterate (usually vialine search).

The acceptable point is the one in trial after return.

void Ipopt::IpoptAlgorithm::AcceptTrialPoint ( ) [private]

Method for accepting the trial point as the new iteration, possibly after adjusting the variable bounds in the NLP.

void Ipopt::IpoptAlgorithm::OutputIteration ( ) [private]

Do all the output for one iteration.

void Ipopt::IpoptAlgorithm::InitializeIterates ( ) [private]

Sets up initial values for the iterates, Corrects the initial values for x and s (force in bounds).

void Ipopt::IpoptAlgorithm::PrintProblemStatistics ( ) [private]

Print the problem size statistics.

void Ipopt::IpoptAlgorithm::ComputeFeasibilityMultipliers ( ) [private]

Compute the Lagrangian multipliers for a feasibility problem.

void Ipopt::IpoptAlgorithm::calc_number_of_bounds	(	const Vector &	x,
		const Vector &	x_L,
		const Vector &	x_U,
		const Matrix &	Px_L,
		const Matrix &	Px_U,
		Index &	n_tot,
		Index &	n_only_lower,
		Index &	n_both,
		Index &	n_only_upper
	)			`[private]`

Method for ensuring that the trial multipliers are not too far from the primal estime.

If a correction is made, new_trial_z is a pointer to the corrected multiplier, and the return value of this method give the magnitutde of the largest correction that we done. If no correction was made, new_trial_z is just a pointer to trial_z, and the return value is zero.

Number Ipopt::IpoptAlgorithm::correct_bound_multiplier	(	const Vector &	trial_z,
		const Vector &	trial_slack,
		const Vector &	trial_compl,
		SmartPtr< const Vector > &	new_trial_z
	)			`[private]`

Method for ensuring that the trial multipliers are not too far from the primal estime.

If a correction is made, new_trial_z is a pointer to the corrected multiplier, and the return value of this method give the magnitutde of the largest correction that we done. If no correction was made, new_trial_z is just a pointer to trial_z, and the return value is zero.