Provides an interface to run trust-region methods for unconstrained optimization algorithms. More...

#include <ROL_TypeU_TrustRegionAlgorithm.hpp>

Inheritance diagram for ROL::TypeU::TrustRegionAlgorithm< Real >:

Public Member Functions
	TrustRegionAlgorithm (ParameterList &parlist, const Ptr< Secant< Real >> &secant=nullPtr)

void	run (Vector< Real > &x, const Vector< Real > &g, Objective< Real > &obj, std::ostream &outStream=std::cout) override
	Run algorithm on unconstrained problems (Type-U). This general interface supports the use of dual optimization vector spaces, where the user does not define the dual() method. More...

void	writeHeader (std::ostream &os) const override
	Print iterate header. More...

void	writeName (std::ostream &os) const override
	Print step name. More...

void	writeOutput (std::ostream &os, const bool print_header=false) const override
	Print iterate status. More...

Public Member Functions inherited from ROL::TypeU::Algorithm< Real >
virtual	~Algorithm ()

	Algorithm ()
	Constructor, given a step and a status test. More...

void	setStatusTest (const Ptr< StatusTest< Real >> &status, bool combineStatus=false)

virtual void	run (Problem< Real > &problem, std::ostream &outStream=std::cout)
	Run algorithm on unconstrained problems (Type-U). This is the primary Type-U interface. More...

virtual void	run (Vector< Real > &x, Objective< Real > &obj, std::ostream &outStream=std::cout)
	Run algorithm on unconstrained problems (Type-U). This is the primary Type-U interface. More...

virtual void	run (Vector< Real > &x, Objective< Real > &obj, Constraint< Real > &linear_con, Vector< Real > &linear_mul, std::ostream &outStream=std::cout)
	Run algorithm on unconstrained problems with explicit linear equality constraints (Type-U). This is the primary Type-U with explicit linear equality constraints interface. More...

virtual void	run (Vector< Real > &x, const Vector< Real > &g, Objective< Real > &obj, Constraint< Real > &linear_con, Vector< Real > &linear_mul, const Vector< Real > &linear_c, std::ostream &outStream=std::cout)
	Run algorithm on unconstrained problems with explicit linear equality constraints (Type-U). This general interface supports the use of dual optimization vector spaces, where the user does not define the dual() method. More...

virtual void	writeExitStatus (std::ostream &os) const

Ptr< const AlgorithmState< Real > >	getState () const

void	reset ()

Private Member Functions
void	initialize (const Vector< Real > &x, const Vector< Real > &g, Vector< Real > &Bg, Objective< Real > &obj, std::ostream &outStream=std::cout)

Real	computeValue (const Vector< Real > &x, Objective< Real > &obj, Real pRed)

void	computeGradient (const Vector< Real > &x, Objective< Real > &obj, bool accept)
	Compute gradient to iteratively satisfy inexactness condition. More...

Private Attributes
Ptr< TrustRegion_U< Real > >	solver_
	Container for trust-region solver object. More...

Ptr< TrustRegionModel_U< Real > >	model_
	Container for trust-region model. More...

ETrustRegionU	etr_
	Trust-region subproblem solver type. More...

Real	delMax_
	Maximum trust-region radius. More...

Real	eta0_
	Step acceptance threshold. More...

Real	eta1_
	Radius decrease threshold. More...

Real	eta2_
	Radius increase threshold. More...

Real	gamma0_
	Radius decrease rate (negative rho). More...

Real	gamma1_
	Radius decrease rate (positive rho). More...

Real	gamma2_
	Radius increase rate. More...

Real	TRsafe_
	Safeguard size for numerically evaluating ratio. More...

Real	eps_
	Safeguard for numerically evaluating ratio. More...

TRUtils::ETRFlag	TRflag_
	Trust-region exit flag. More...

int	SPflag_
	Subproblem solver termination flag. More...

int	SPiter_
	Subproblem solver iteration count. More...

bool	useNM_

int	NMstorage_

ESecant	esec_
	Secant type. More...

bool	useSecantPrecond_

bool	useSecantHessVec_

std::vector< bool >	useInexact_
	Flags for inexact (0) objective function, (1) gradient, (2) Hessian. More...

Real	scale0_
	Scale for inexact gradient computation. More...

Real	scale1_
	Scale for inexact gradient computation. More...

Real	scale_

Real	omega_

Real	force_

Real	forceFactor_

int	updateIter_

Real	gtol_

int	verbosity_
	Print additional information to screen if > 0. More...

bool	printHeader_
	Print header at every iteration. More...

Additional Inherited Members
Protected Member Functions inherited from ROL::TypeU::Algorithm< Real >
void	initialize (const Vector< Real > &x, const Vector< Real > &g)

Protected Attributes inherited from ROL::TypeU::Algorithm< Real >
const Ptr< CombinedStatusTest < Real > >	status_

const Ptr< AlgorithmState< Real > >	state_

Detailed Description

template<typename Real>
class ROL::TypeU::TrustRegionAlgorithm< Real >

Provides an interface to run trust-region methods for unconstrained optimization algorithms.

Definition at line 62 of file ROL_TypeU_TrustRegionAlgorithm.hpp.

Constructor & Destructor Documentation

template<typename Real >

ROL::TypeU::TrustRegionAlgorithm< Real >::TrustRegionAlgorithm	(	ParameterList &	parlist,
		const Ptr< Secant< Real >> &	secant = `nullPtr`
	)

Definition at line 53 of file ROL_TypeU_TrustRegionAlgorithm_Def.hpp.

Member Function Documentation

template<typename Real >

void ROL::TypeU::TrustRegionAlgorithm< Real >::run	(	Vector< Real > &	x,
		const Vector< Real > &	g,
		Objective< Real > &	obj,
		std::ostream &	outStream = `std::cout`
	)

overridevirtual

Run algorithm on unconstrained problems (Type-U). This general interface supports the use of dual optimization vector spaces, where the user does not define the dual() method.

Implements ROL::TypeU::Algorithm< Real >.

Definition at line 189 of file ROL_TypeU_TrustRegionAlgorithm_Def.hpp.

References ROL::Accept, ROL::Vector< Real >::clone(), ROL::TRUtils::NPOSPREDNEG, ROL::Vector< Real >::plus(), ROL::TRUtils::POSPREDNEG, ROL::Revert, ROL::Vector< Real >::set(), ROL::TRUtils::SUCCESS, ROL::TRUtils::TRNAN, ROL::Objective< Real >::update(), ROL::TypeU::Algorithm< Real >::writeExitStatus(), and zero.

Referenced by main().

template<typename Real >

void ROL::TypeU::TrustRegionAlgorithm< Real >::writeHeader ( std::ostream & os ) const

overridevirtual

Print iterate header.

Reimplemented from ROL::TypeU::Algorithm< Real >.

Definition at line 274 of file ROL_TypeU_TrustRegionAlgorithm_Def.hpp.

References ROL::CG_FLAG_SUCCESS, ROL::CG_FLAG_UNDEFINED, ROL::ECGFlagToString(), ROL::TRUtils::ETRFlagToString(), ROL::NumberToString(), ROL::TRUtils::SUCCESS, ROL::TRUSTREGION_U_SPG, ROL::TRUSTREGION_U_TRUNCATEDCG, and ROL::TRUtils::UNDEFINED.

template<typename Real >

void ROL::TypeU::TrustRegionAlgorithm< Real >::writeName ( std::ostream & os ) const

overridevirtual

Print step name.

Reimplemented from ROL::TypeU::Algorithm< Real >.

Definition at line 330 of file ROL_TypeU_TrustRegionAlgorithm_Def.hpp.

References ROL::ESecantToString(), and ROL::ETrustRegionUToString().

template<typename Real >

void ROL::TypeU::TrustRegionAlgorithm< Real >::writeOutput	(	std::ostream &	os,
		const bool	write_header = `false`
	)		const

overridevirtual

Print iterate status.

Reimplemented from ROL::TypeU::Algorithm< Real >.

Definition at line 351 of file ROL_TypeU_TrustRegionAlgorithm_Def.hpp.

References ROL::TRUSTREGION_U_SPG, and ROL::TRUSTREGION_U_TRUNCATEDCG.

template<typename Real >

void ROL::TypeU::TrustRegionAlgorithm< Real >::initialize	(	const Vector< Real > &	x,
		const Vector< Real > &	g,
		Vector< Real > &	Bg,
		Objective< Real > &	obj,
		std::ostream &	outStream = `std::cout`
	)

private

Definition at line 109 of file ROL_TypeU_TrustRegionAlgorithm_Def.hpp.

References ROL::Initial, ROL::TypeU::Algorithm< Real >::initialize(), ROL::Vector< Real >::norm(), ROL::Objective< Real >::update(), ROL::Objective< Real >::value(), and zero.

template<typename Real >

Real ROL::TypeU::TrustRegionAlgorithm< Real >::computeValue	(	const Vector< Real > &	x,
		Objective< Real > &	obj,
		Real	pRed
	)

private

Definition at line 143 of file ROL_TypeU_TrustRegionAlgorithm_Def.hpp.

References ROL::Trial, ROL::Objective< Real >::update(), and ROL::Objective< Real >::value().

template<typename Real >

void ROL::TypeU::TrustRegionAlgorithm< Real >::computeGradient	(	const Vector< Real > &	x,
		Objective< Real > &	obj,
		bool	accept
	)

private

Compute gradient to iteratively satisfy inexactness condition.

This function attempts to ensure that the inexact gradient condition,

\[ \|g_k-\nabla J(x_k)\|_{\mathcal{X}} \le \kappa_1\min\{\,\|g_k\|_{\mathcal{X}},\,\Delta_k\,\}, \]

is satisfied. This function works under the assumption that the gradient function returns a gradient approximation which satisfies the error tolerance prescribed by the tol input parameter.

Parameters