Ipopt::PDFullSpaceSolver Class Reference

This is the implemetation of the Primal-Dual System, using the full space approach with a direct linear solver. More...

#include <IpPDFullSpaceSolver.hpp>

Inheritance diagram for Ipopt::PDFullSpaceSolver:

[legend]

Collaboration diagram for Ipopt::PDFullSpaceSolver:

[legend]

List of all members.

Public Member Functions
bool	InitializeImpl (const OptionsList &options, const std::string &prefix)
	overloaded from AlgorithmStrategyObject
virtual bool	Solve (Number alpha, Number beta, const IteratesVector &rhs, IteratesVector &res, bool allow_inexact=false, bool improve_solution=false)
	Solve the primal dual system, given one right hand side.
/Destructor

	PDFullSpaceSolver (AugSystemSolver &augSysSolver, PDPerturbationHandler &perturbHandler)
	Constructor that takes in the Augmented System solver that is to be used inside.
virtual	~PDFullSpaceSolver ()
	Default destructor.
Static Public Member Functions


static void	RegisterOptions (SmartPtr< RegisteredOptions > roptions)
	Methods for IpoptType.
Private Member Functions
bool	SolveOnce (bool resolve_unmodified, bool pretend_singular, const SymMatrix &W, const Matrix &J_c, const Matrix &J_d, const Matrix &Px_L, const Matrix &Px_U, const Matrix &Pd_L, const Matrix &Pd_U, const Vector &z_L, const Vector &z_U, const Vector &v_L, const Vector &v_U, const Vector &slack_x_L, const Vector &slack_x_U, const Vector &slack_s_L, const Vector &slack_s_U, const Vector &sigma_x, const Vector &sigma_s, Number alpha, Number beta, const IteratesVector &rhs, IteratesVector &res)
	Internal function for a single backsolve (which will be used for iterative refinement on the outside).
void	ComputeResiduals (const SymMatrix &W, const Matrix &J_c, const Matrix &J_d, const Matrix &Px_L, const Matrix &Px_U, const Matrix &Pd_L, const Matrix &Pd_U, const Vector &z_L, const Vector &z_U, const Vector &v_L, const Vector &v_U, const Vector &slack_x_L, const Vector &slack_x_U, const Vector &slack_s_L, const Vector &slack_s_U, const Vector &sigma_x, const Vector &sigma_s, Number alpha, Number beta, const IteratesVector &rhs, const IteratesVector &res, IteratesVector &resid)
	Internal function for computing the residual (resid) given the right hand side (rhs) and the solution of the system (res).
Number	ComputeResidualRatio (const IteratesVector &rhs, const IteratesVector &res, const IteratesVector &resid)
	Internal function for computing the ratio of the residual compared to the right hand side and solution.
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.
	PDFullSpaceSolver ()
	Default Constructor.
PDFullSpaceSolver &	operator= (const PDFullSpaceSolver &)
	Overloaded Equals Operator.
Auxilliary functions

void	SinvBlrmZPTdBr (Number alpha, const Vector &S, const Vector &R, const Vector &Z, const Matrix &P, const Vector &g, Vector &X)
	Compute $x = S^{-1}(r + \alpha Z P^T d)$ .
Private Attributes
Strategy objects to hold on to.

SmartPtr< AugSystemSolver >	augSysSolver_
	Pointer to the Solver for the augmented system.
SmartPtr< PDPerturbationHandler >	perturbHandler_
	Pointer to the Perturbation Handler.
Data about the correction made to the system

CachedResults< void * >	dummy_cache_
	A dummy cache to figure out if the deltas are still up to date.
bool	augsys_improved_
	Flag indicating if for the current matrix the solution quality of the augmented system solver has already been increased.
Parameters

Index	min_refinement_steps_
	Minimal number of iterative refinement performed per backsolve.
Index	max_refinement_steps_
	Maximal number of iterative refinement performed per backsolve.
Number	residual_ratio_max_
	Maximal allowed ratio of the norm of the residual over the norm of the right hand side and solution.
Number	residual_ratio_singular_
	If the residual_ratio is larger than this value after trying to improve the solution, the linear system is assumed to be singular and modified.
Number	residual_improvement_factor_
	Factor defining require improvement to consider iterative refinement successful.
Number	neg_curv_test_tol_
	Tolernace for heuristic to ignore wrong inertia.

Detailed Description

This is the implemetation of the Primal-Dual System, using the full space approach with a direct linear solver.

A note on the iterative refinement: We perform at least min_refinement_steps number of iterative refinement steps. If after one iterative refinement the quality of the solution (defined in ResidualRatio) does not improve or the maximal number of iterative refinement steps is exceeded before the tolerance residual_ratio_max_ is satisfied, we first ask the linear solver to solve the system more accurately (e.g. by increasing the pivot tolerance). If that doesn't help or is not possible, we treat the system, as if it is singular (i.e. increase delta's).

Definition at line 32 of file IpPDFullSpaceSolver.hpp.

Constructor & Destructor Documentation

Ipopt::PDFullSpaceSolver::PDFullSpaceSolver	(	AugSystemSolver &	augSysSolver,
		PDPerturbationHandler &	perturbHandler
	)

Constructor that takes in the Augmented System solver that is to be used inside.

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

Default destructor.

Ipopt::PDFullSpaceSolver::PDFullSpaceSolver ( ) [private]

Default Constructor.

Member Function Documentation

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

overloaded from AlgorithmStrategyObject

Implements Ipopt::PDSystemSolver.

virtual bool Ipopt::PDFullSpaceSolver::Solve	(	Number	alpha,
		Number	beta,
		const IteratesVector &	rhs,
		IteratesVector &	res,
		bool	allow_inexact = `false`,
		bool	improve_solution = `false`
	)			`[virtual]`

Solve the primal dual system, given one right hand side.

Implements Ipopt::PDSystemSolver.

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

Methods for IpoptType.

PDFullSpaceSolver& Ipopt::PDFullSpaceSolver::operator= ( const PDFullSpaceSolver & ) [private]

Overloaded Equals Operator.

Reimplemented from Ipopt::PDSystemSolver.

bool Ipopt::PDFullSpaceSolver::SolveOnce	(	bool	resolve_unmodified,
		bool	pretend_singular,
		const SymMatrix &	W,
		const Matrix &	J_c,
		const Matrix &	J_d,
		const Matrix &	Px_L,
		const Matrix &	Px_U,
		const Matrix &	Pd_L,
		const Matrix &	Pd_U,
		const Vector &	z_L,
		const Vector &	z_U,
		const Vector &	v_L,
		const Vector &	v_U,
		const Vector &	slack_x_L,
		const Vector &	slack_x_U,
		const Vector &	slack_s_L,
		const Vector &	slack_s_U,
		const Vector &	sigma_x,
		const Vector &	sigma_s,
		Number	alpha,
		Number	beta,
		const IteratesVector &	rhs,
		IteratesVector &	res
	)			`[private]`

Internal function for a single backsolve (which will be used for iterative refinement on the outside).

This method returns false, if for some reason the linear system could not be solved (e.g. when the regularization parameter becomes too large.)

void Ipopt::PDFullSpaceSolver::ComputeResiduals	(	const SymMatrix &	W,
		const Matrix &	J_c,
		const Matrix &	J_d,
		const Matrix &	Px_L,
		const Matrix &	Px_U,
		const Matrix &	Pd_L,
		const Matrix &	Pd_U,
		const Vector &	z_L,
		const Vector &	z_U,
		const Vector &	v_L,
		const Vector &	v_U,
		const Vector &	slack_x_L,
		const Vector &	slack_x_U,
		const Vector &	slack_s_L,
		const Vector &	slack_s_U,
		const Vector &	sigma_x,
		const Vector &	sigma_s,
		Number	alpha,
		Number	beta,
		const IteratesVector &	rhs,
		const IteratesVector &	res,
		IteratesVector &	resid
	)			`[private]`

Internal function for computing the residual (resid) given the right hand side (rhs) and the solution of the system (res).

Number Ipopt::PDFullSpaceSolver::ComputeResidualRatio	(	const IteratesVector &	rhs,
		const IteratesVector &	res,
		const IteratesVector &	resid
	)			`[private]`

Internal function for computing the ratio of the residual compared to the right hand side and solution.

The smaller this value, the better the solution.

void Ipopt::PDFullSpaceSolver::SinvBlrmZPTdBr	(	Number	alpha,
		const Vector &	S,
		const Vector &	R,
		const Vector &	Z,
		const Matrix &	P,
		const Vector &	g,
		Vector &	X
	)			`[private]`