Solves Quadratic Programming (QP) problems using QPSchur. More...

#include <ConstrainedOptPack_QPSolverRelaxedQPSchur.hpp>

Inheritance diagram for ConstrainedOptPack::QPSolverRelaxedQPSchur:

Classes
class	InitKKTSystem
	Interface for the object that forms the initial KKT system {abstract}. More...

class	ReinitKKTSystem
	Interface for the object that can reform an initial KKT system dynamically {abstract}. More...

Public Types
enum	ELocalOutputLevel { USE_INPUT_ARG = -1, NO_OUTPUT = 0, OUTPUT_BASIC_INFO = 1, OUTPUT_ITER_SUMMARY = 2, OUTPUT_ITER_STEPS = 3, OUTPUT_ACT_SET = 4, OUTPUT_ITER_QUANTITIES = 5 }
	Output level. More...

Public Types inherited from ConstrainedOptPack::QPSolverRelaxed
enum	EOutputLevel { PRINT_NONE = 0, PRINT_BASIC_INFO = 1, PRINT_ITER_SUMMARY = 2, PRINT_ITER_STEPS = 3, PRINT_ITER_ACT_SET = 4, PRINT_ITER_VECTORS = 5, PRINT_EVERY_THING = 6 }
	Enumeration for the amount of output to create from `solve_qp()`. More...

enum	ERunTests { RUN_TESTS, NO_TESTS }
	Enumeration for if to run internal tests or not. More...

Public Member Functions
	STANDARD_COMPOSITION_MEMBERS (InitKKTSystem, init_kkt_sys)
	Strategy object that sets up the initial KKT system. More...

	STANDARD_COMPOSITION_MEMBERS (QPSchurPack::ConstraintsRelaxedStd, constraints)
	Constraints object. More...

	STANDARD_MEMBER_COMPOSITION_MEMBERS (value_type, max_qp_iter_frac)
	Set the maximum number of QP iterations as `max_itr = max_qp_iter_frac * n`. More...

	STANDARD_MEMBER_COMPOSITION_MEMBERS (value_type, max_real_runtime)
	Set the maximum real run-time in minutes. More...

	STANDARD_MEMBER_COMPOSITION_MEMBERS (QPSchurPack::ConstraintsRelaxedStd::EInequalityPickPolicy, inequality_pick_policy)
	Policy used to select a violated constraint. More...

	STANDARD_MEMBER_COMPOSITION_MEMBERS (ELocalOutputLevel, print_level)
	Set the output level for QPSchur. More...

	STANDARD_MEMBER_COMPOSITION_MEMBERS (value_type, bounds_tol)
	Set the feasibility tolerance for the bound constriants. More...

	STANDARD_MEMBER_COMPOSITION_MEMBERS (value_type, inequality_tol)
	Set the feasibility tolerance for the general inequality constraints. More...

	STANDARD_MEMBER_COMPOSITION_MEMBERS (value_type, equality_tol)
	Set the feasibility tolerance for the general equality constriants. More...

	STANDARD_MEMBER_COMPOSITION_MEMBERS (value_type, loose_feas_tol)
	Set a looser feasibility tolerance ( > feas_tol ) More...

	STANDARD_MEMBER_COMPOSITION_MEMBERS (value_type, dual_infeas_tol)
	Set the tolerence where a scaled Langrange multiplier is considered degenerate. More...

	STANDARD_MEMBER_COMPOSITION_MEMBERS (value_type, huge_primal_step)
	Set the tolerence for the size of the step in the primal space that is considered to be a near infinite step. This is used to determine if the KKT system is near singular. More...

	STANDARD_MEMBER_COMPOSITION_MEMBERS (value_type, huge_dual_step)
	Set the tolerence for the size of the step in the dual space that is considered to be a near infinite step. This is used to determine if the constriants are infeasible. More...

	STANDARD_MEMBER_COMPOSITION_MEMBERS (value_type, bigM)
	<<std member="" comp>="">> members for the Big M parameter used in the objective. More...

	STANDARD_MEMBER_COMPOSITION_MEMBERS (value_type, warning_tol)
	<<std member="" comp>="">> members for the warning tolerance for tests. More...

	STANDARD_MEMBER_COMPOSITION_MEMBERS (value_type, error_tol)
	<<std member="" comp>="">> members for the error tolerance for tests. More...

	STANDARD_MEMBER_COMPOSITION_MEMBERS (size_type, iter_refine_min_iter)
	Set the minimum number of refinement iterations to perform when using iterative refinement. More...

	STANDARD_MEMBER_COMPOSITION_MEMBERS (size_type, iter_refine_max_iter)
	Set the maximum number of refinement iterations to perform when using iterative refinement. More...

	STANDARD_MEMBER_COMPOSITION_MEMBERS (value_type, iter_refine_opt_tol)
	Set the maxinum scaled tolerance the residual of the optimality conditions must be before terminating iterative refinement. More...

	STANDARD_MEMBER_COMPOSITION_MEMBERS (value_type, iter_refine_feas_tol)
	Set the maxinum scaled tolerance the residual of the feasibility conditions must be before terminating iterative refinement. More...

	STANDARD_MEMBER_COMPOSITION_MEMBERS (bool, iter_refine_at_solution)
	Set whether iterative refinement is automatically used once the solution is found. More...

	STANDARD_MEMBER_COMPOSITION_MEMBERS (value_type, pivot_warning_tol)
	Set the relative tolerance for pivots in the schur complement under which a waning will be printed (see MatrixSymAddDelUpdateable) for near singular updates. More...

	STANDARD_MEMBER_COMPOSITION_MEMBERS (value_type, pivot_singular_tol)
	Set the relative tolerance for pivots in the schur complement under which a singularity exception will be thrown (see MatrixSymAddDelUpdateable) for singular updates. More...

	STANDARD_MEMBER_COMPOSITION_MEMBERS (value_type, pivot_wrong_inertia_tol)
	Set the relative tolerance for pivots in the schur complement over which a wrong inertia exception will be throw (see MatrixSymAddDelUpdateable) for updates with the wrong inertia. More...

	STANDARD_MEMBER_COMPOSITION_MEMBERS (bool, add_equalities_initially)
	Set whether equality constriants are to be added to the active set initialy to the schur complement or not. More...

	QPSolverRelaxedQPSchur (const init_kkt_sys_ptr_t &init_kkt_sys=Teuchos::null, const constraints_ptr_t &constraints=Teuchos::rcp(new QPSchurPack::ConstraintsRelaxedStd), value_type max_qp_iter_frac=10.0, value_type max_real_runtime=1e+20, QPSchurPack::ConstraintsRelaxedStd::EInequalityPickPolicy inequality_pick_policy=QPSchurPack::ConstraintsRelaxedStd::ADD_BOUNDS_THEN_MOST_VIOLATED_INEQUALITY, ELocalOutputLevel print_level=USE_INPUT_ARG, value_type bounds_tol=-1.0, value_type inequality_tol=-1.0, value_type equality_tol=-1.0, value_type loose_feas_tol=-1.0, value_type dual_infeas_tol=-1.0, value_type huge_primal_step=-1.0, value_type huge_dual_step=-1.0, value_type bigM=1e+10, value_type warning_tol=1e-10, value_type error_tol=1e-5, size_type iter_refine_min_iter=1, size_type iter_refine_max_iter=3, value_type iter_refine_opt_tol=1e-12, value_type iter_refine_feas_tol=1e-12, bool iter_refine_at_solution=true, value_type pivot_warning_tol=1e-8, value_type pivot_singular_tol=1e-11, value_type pivot_wrong_inertia_tol=1e-11, bool add_equalities_initially=true)

	~QPSolverRelaxedQPSchur ()

Public Member Functions inherited from ConstrainedOptPack::QPSolverRelaxed
	STANDARD_MEMBER_COMPOSITION_MEMBERS (value_type, infinite_bound)
	Set the scalar that will be used to identigy infinite bounds. More...

	QPSolverRelaxed ()

virtual	~QPSolverRelaxed ()

virtual QPSolverStats::ESolutionType	solve_qp (std::ostream out, EOutputLevel olevel, ERunTests test_what, const Vector &g, const MatrixSymOp &G, value_type etaL, const Vector &dL, const Vector &dU, const MatrixOp &E, BLAS_Cpp::Transp trans_E, const Vector &b, const Vector &eL, const Vector &eU, const MatrixOp &F, BLAS_Cpp::Transp trans_F, const Vector &f, value_type obj_d, value_type eta, VectorMutable d, VectorMutable nu, VectorMutable mu, VectorMutable Ed, VectorMutable lambda, VectorMutable *Fd)
	Solve the QP. More...

virtual QPSolverStats::ESolutionType	solve_qp (std::ostream out, EOutputLevel olevel, ERunTests test_what, const Vector &g, const MatrixSymOp &G, value_type etaL, const Vector &dL, const Vector &dU, const MatrixOp &E, BLAS_Cpp::Transp trans_E, const Vector &b, const Vector &eL, const Vector &eU, value_type obj_d, value_type eta, VectorMutable d, VectorMutable nu, VectorMutable mu, VectorMutable *Ed)
	Solve the QP without general equality constrants. More...

virtual QPSolverStats::ESolutionType	solve_qp (std::ostream out, EOutputLevel olevel, ERunTests test_what, const Vector &g, const MatrixSymOp &G, value_type etaL, const Vector &dL, const Vector &dU, const MatrixOp &F, BLAS_Cpp::Transp trans_F, const Vector &f, value_type obj_d, value_type eta, VectorMutable d, VectorMutable nu, VectorMutable lambda, VectorMutable *Fd)
	Solve the QP without general inequality constrants. More...

virtual QPSolverStats::ESolutionType	solve_qp (std::ostream out, EOutputLevel olevel, ERunTests test_what, const Vector &g, const MatrixSymOp &G, const Vector &dL, const Vector &dU, value_type obj_d, VectorMutable d, VectorMutable nu)
	Solve the QP without general equality or inequality constrants (no relaxation needed). More...

virtual QPSolverStats::ESolutionType	solve_qp (std::ostream out, EOutputLevel olevel, ERunTests test_what, const Vector &g, const MatrixSymOp &G, value_type etaL, const Vector dL, const Vector dU, const MatrixOp E, BLAS_Cpp::Transp trans_E, const Vector b, const Vector eL, const Vector eU, const MatrixOp F, BLAS_Cpp::Transp trans_F, const Vector f, value_type obj_d, value_type eta, VectorMutable d, VectorMutable nu, VectorMutable mu, VectorMutable Ed, VectorMutable lambda, VectorMutable *Fd)
	This is a more flexible function where the client can set different constraints to be included. More...

Private Attributes
QPSolverStats	qp_stats_

QPSchur	qp_solver_

QPSchurPack::QPInitFixedFreeStd	qp_

MatrixSymHessianRelaxNonSing	G_relaxed_

VectorMutableDense	bigM_vec_

MatrixSymAddDelBunchKaufman	schur_comp_

DVector	g_relaxed_

DVector	b_X_

InitKKTSystem::Ko_ptr_t	Ko_

DVector	fo_

Overridden from QPSolverRelaxed
QPSolverStats	get_qp_stats () const

void	release_memory ()

QPSolverStats::ESolutionType	imp_solve_qp (std::ostream out, EOutputLevel olevel, ERunTests test_what, const Vector &g, const MatrixSymOp &G, value_type etaL, const Vector dL, const Vector dU, const MatrixOp E, BLAS_Cpp::Transp trans_E, const Vector b, const Vector eL, const Vector eU, const MatrixOp F, BLAS_Cpp::Transp trans_F, const Vector f, value_type obj_d, value_type eta, VectorMutable d, VectorMutable nu, VectorMutable mu, VectorMutable Ed, VectorMutable lambda, VectorMutable *Fd)

Detailed Description

Solves Quadratic Programming (QP) problems using QPSchur.

This is the only subclass needed for QPSchur. All of the specifics of how the initial KKT system is formed is delegated to a strategy object of type InitKKTSystem (see below).

Definition at line 63 of file ConstrainedOptPack_QPSolverRelaxedQPSchur.hpp.

Member Enumeration Documentation

enum ConstrainedOptPack::QPSolverRelaxedQPSchur::ELocalOutputLevel

Output level.

Enumerator
USE_INPUT_ARG
NO_OUTPUT
OUTPUT_BASIC_INFO
OUTPUT_ITER_SUMMARY
OUTPUT_ITER_STEPS
OUTPUT_ACT_SET
OUTPUT_ITER_QUANTITIES

Definition at line 220 of file ConstrainedOptPack_QPSolverRelaxedQPSchur.hpp.

Constructor & Destructor Documentation

ConstrainedOptPack::QPSolverRelaxedQPSchur::QPSolverRelaxedQPSchur	(	const init_kkt_sys_ptr_t &	init_kkt_sys = `Teuchos::null`,
		const constraints_ptr_t &	constraints = `Teuchos::rcp(new QPSchurPack::ConstraintsRelaxedStd)`,
		value_type	max_qp_iter_frac = `10.0`,
		value_type	max_real_runtime = `1e+20`,
		QPSchurPack::ConstraintsRelaxedStd::EInequalityPickPolicy	inequality_pick_policy = `QPSchurPack::ConstraintsRelaxedStd::ADD_BOUNDS_THEN_MOST_VIOLATED_INEQUALITY`,
		ELocalOutputLevel	print_level = `USE_INPUT_ARG`,
		value_type	bounds_tol = `-1.0`,
		value_type	inequality_tol = `-1.0`,
		value_type	equality_tol = `-1.0`,
		value_type	loose_feas_tol = `-1.0`,
		value_type	dual_infeas_tol = `-1.0`,
		value_type	huge_primal_step = `-1.0`,
		value_type	huge_dual_step = `-1.0`,
		value_type	bigM = `1e+10`,
		value_type	warning_tol = `1e-10`,
		value_type	error_tol = `1e-5`,
		size_type	iter_refine_min_iter = `1`,
		size_type	iter_refine_max_iter = `3`,
		value_type	iter_refine_opt_tol = `1e-12`,
		value_type	iter_refine_feas_tol = `1e-12`,
		bool	iter_refine_at_solution = `true`,
		value_type	pivot_warning_tol = `1e-8`,
		value_type	pivot_singular_tol = `1e-11`,
		value_type	pivot_wrong_inertia_tol = `1e-11`,
		bool	add_equalities_initially = `true`
	)

Definition at line 60 of file ConstrainedOptPack_QPSolverRelaxedQPSchur.cpp.

ConstrainedOptPack::QPSolverRelaxedQPSchur::~QPSolverRelaxedQPSchur ( )

Definition at line 115 of file ConstrainedOptPack_QPSolverRelaxedQPSchur.cpp.

Member Function Documentation

ConstrainedOptPack::QPSolverRelaxedQPSchur::STANDARD_COMPOSITION_MEMBERS	(	InitKKTSystem	,
		init_kkt_sys
	)

Strategy object that sets up the initial KKT system.

ConstrainedOptPack::QPSolverRelaxedQPSchur::STANDARD_COMPOSITION_MEMBERS	(	QPSchurPack::ConstraintsRelaxedStd	,
		constraints
	)

Constraints object.

ConstrainedOptPack::QPSolverRelaxedQPSchur::STANDARD_MEMBER_COMPOSITION_MEMBERS	(	value_type	,
		max_qp_iter_frac
	)

Set the maximum number of QP iterations as max_itr = max_qp_iter_frac * n.

ConstrainedOptPack::QPSolverRelaxedQPSchur::STANDARD_MEMBER_COMPOSITION_MEMBERS	(	value_type	,
		max_real_runtime
	)

Set the maximum real run-time in minutes.

ConstrainedOptPack::QPSolverRelaxedQPSchur::STANDARD_MEMBER_COMPOSITION_MEMBERS	(	QPSchurPack::ConstraintsRelaxedStd::EInequalityPickPolicy	,
		inequality_pick_policy
	)

Policy used to select a violated constraint.

ConstrainedOptPack::QPSolverRelaxedQPSchur::STANDARD_MEMBER_COMPOSITION_MEMBERS	(	ELocalOutputLevel	,
		print_level
	)

Set the output level for QPSchur.

ConstrainedOptPack::QPSolverRelaxedQPSchur::STANDARD_MEMBER_COMPOSITION_MEMBERS	(	value_type	,
		bounds_tol
	)

Set the feasibility tolerance for the bound constriants.

ConstrainedOptPack::QPSolverRelaxedQPSchur::STANDARD_MEMBER_COMPOSITION_MEMBERS	(	value_type	,
		inequality_tol
	)

Set the feasibility tolerance for the general inequality constraints.

ConstrainedOptPack::QPSolverRelaxedQPSchur::STANDARD_MEMBER_COMPOSITION_MEMBERS	(	value_type	,
		equality_tol
	)

Set the feasibility tolerance for the general equality constriants.

ConstrainedOptPack::QPSolverRelaxedQPSchur::STANDARD_MEMBER_COMPOSITION_MEMBERS	(	value_type	,
		loose_feas_tol
	)

Set a looser feasibility tolerance ( > feas_tol )

ConstrainedOptPack::QPSolverRelaxedQPSchur::STANDARD_MEMBER_COMPOSITION_MEMBERS	(	value_type	,
		dual_infeas_tol
	)

Set the tolerence where a scaled Langrange multiplier is considered degenerate.

ConstrainedOptPack::QPSolverRelaxedQPSchur::STANDARD_MEMBER_COMPOSITION_MEMBERS	(	value_type	,
		huge_primal_step
	)

Set the tolerence for the size of the step in the primal space that is considered to be a near infinite step. This is used to determine if the KKT system is near singular.

ConstrainedOptPack::QPSolverRelaxedQPSchur::STANDARD_MEMBER_COMPOSITION_MEMBERS	(	value_type	,
		huge_dual_step
	)

Set the tolerence for the size of the step in the dual space that is considered to be a near infinite step. This is used to determine if the constriants are infeasible.

ConstrainedOptPack::QPSolverRelaxedQPSchur::STANDARD_MEMBER_COMPOSITION_MEMBERS	(	value_type	,
		bigM
	)

<<std member="" comp>="">> members for the Big M parameter used in the objective.

ConstrainedOptPack::QPSolverRelaxedQPSchur::STANDARD_MEMBER_COMPOSITION_MEMBERS	(	value_type	,
		warning_tol
	)

<<std member="" comp>="">> members for the warning tolerance for tests.

ConstrainedOptPack::QPSolverRelaxedQPSchur::STANDARD_MEMBER_COMPOSITION_MEMBERS	(	value_type	,
		error_tol
	)

<<std member="" comp>="">> members for the error tolerance for tests.

ConstrainedOptPack::QPSolverRelaxedQPSchur::STANDARD_MEMBER_COMPOSITION_MEMBERS	(	size_type	,
		iter_refine_min_iter
	)

Set the minimum number of refinement iterations to perform when using iterative refinement.

ConstrainedOptPack::QPSolverRelaxedQPSchur::STANDARD_MEMBER_COMPOSITION_MEMBERS	(	size_type	,
		iter_refine_max_iter
	)

Set the maximum number of refinement iterations to perform when using iterative refinement.

ConstrainedOptPack::QPSolverRelaxedQPSchur::STANDARD_MEMBER_COMPOSITION_MEMBERS	(	value_type	,
		iter_refine_opt_tol
	)

Set the maxinum scaled tolerance the residual of the optimality conditions must be before terminating iterative refinement.

ConstrainedOptPack::QPSolverRelaxedQPSchur::STANDARD_MEMBER_COMPOSITION_MEMBERS	(	value_type	,
		iter_refine_feas_tol
	)

Set the maxinum scaled tolerance the residual of the feasibility conditions must be before terminating iterative refinement.

ConstrainedOptPack::QPSolverRelaxedQPSchur::STANDARD_MEMBER_COMPOSITION_MEMBERS	(	bool	,
		iter_refine_at_solution
	)

Set whether iterative refinement is automatically used once the solution is found.

ConstrainedOptPack::QPSolverRelaxedQPSchur::STANDARD_MEMBER_COMPOSITION_MEMBERS	(	value_type	,
		pivot_warning_tol
	)

Set the relative tolerance for pivots in the schur complement under which a waning will be printed (see MatrixSymAddDelUpdateable) for near singular updates.

ConstrainedOptPack::QPSolverRelaxedQPSchur::STANDARD_MEMBER_COMPOSITION_MEMBERS	(	value_type	,
		pivot_singular_tol
	)

Set the relative tolerance for pivots in the schur complement under which a singularity exception will be thrown (see MatrixSymAddDelUpdateable) for singular updates.

ConstrainedOptPack::QPSolverRelaxedQPSchur::STANDARD_MEMBER_COMPOSITION_MEMBERS	(	value_type	,
		pivot_wrong_inertia_tol
	)

Set the relative tolerance for pivots in the schur complement over which a wrong inertia exception will be throw (see MatrixSymAddDelUpdateable) for updates with the wrong inertia.

ConstrainedOptPack::QPSolverRelaxedQPSchur::STANDARD_MEMBER_COMPOSITION_MEMBERS	(	bool	,
		add_equalities_initially
	)

Set whether equality constriants are to be added to the active set initialy to the schur complement or not.

QPSolverStats ConstrainedOptPack::QPSolverRelaxedQPSchur::get_qp_stats ( ) const

virtual

Implements ConstrainedOptPack::QPSolverRelaxed.

Definition at line 123 of file ConstrainedOptPack_QPSolverRelaxedQPSchur.cpp.

void ConstrainedOptPack::QPSolverRelaxedQPSchur::release_memory ( )

virtual

Implements ConstrainedOptPack::QPSolverRelaxed.

Definition at line 128 of file ConstrainedOptPack_QPSolverRelaxedQPSchur.cpp.

QPSolverStats::ESolutionType ConstrainedOptPack::QPSolverRelaxedQPSchur::imp_solve_qp	(	std::ostream *	out,
		EOutputLevel	olevel,
		ERunTests	test_what,
		const Vector &	g,
		const MatrixSymOp &	G,
		value_type	etaL,
		const Vector *	dL,
		const Vector *	dU,
		const MatrixOp *	E,
		BLAS_Cpp::Transp	trans_E,
		const Vector *	b,
		const Vector *	eL,
		const Vector *	eU,
		const MatrixOp *	F,
		BLAS_Cpp::Transp	trans_F,
		const Vector *	f,
		value_type *	obj_d,
		value_type *	eta,
		VectorMutable *	d,
		VectorMutable *	nu,
		VectorMutable *	mu,
		VectorMutable *	Ed,
		VectorMutable *	lambda,
		VectorMutable *	Fd
	)