Provides the interface to apply a linear operator. More...

#include <ROL_LinearOperator.hpp>

Inheritance diagram for ROL::LinearOperator< Real >:

Public Member Functions
virtual	~LinearOperator ()

virtual void	update (const Vector< Real > &x, bool flag=true, int iter=-1)
	Update linear operator. More...

virtual void	apply (Vector< Real > &Hv, const Vector< Real > &v, Real &tol) const =0
	Apply linear operator. More...

virtual void	applyInverse (Vector< Real > &Hv, const Vector< Real > &v, Real &tol) const
	Apply inverse of linear operator. More...

virtual void	applyAdjoint (Vector< Real > &Hv, const Vector< Real > &v, Real &tol) const
	Apply adjoint of linear operator. More...

virtual void	applyAdjointInverse (Vector< Real > &Hv, const Vector< Real > &v, Real &tol) const
	Apply adjoint of the inverse linear operator. More...

Detailed Description

template<class Real>
class ROL::LinearOperator< Real >

Provides the interface to apply a linear operator.

ROL's linear operator interface is designed to interface with ROL's Krylov methods. These linear operators often represent projected Hessians or preconditioners. The basic operator interace, to be implemented by the user, requires:

apply – apply operator to a vector.

The user may also implement:

update – update the state of the linear operator.
applyInverse – apply the inverse operator to a vector.
applyAdjoint – apply the adjoint of the operator to a vector.
applyAdjointInverse – apply the adjoint of the inverse operator to a vector.

Definition at line 37 of file ROL_LinearOperator.hpp.

Constructor & Destructor Documentation

template<class Real>

virtual ROL::LinearOperator< Real >::~LinearOperator ( )

inlinevirtual

Definition at line 40 of file ROL_LinearOperator.hpp.

Member Function Documentation

template<class Real>

virtual void ROL::LinearOperator< Real >::update	(	const Vector< Real > &	x,
		bool	flag = `true`,
		int	iter = `-1`
	)

inlinevirtual

Update linear operator.

This function updates the linear operator at new iterations.

Parameters

[in]	x	is the new iterate.
[in]	flag	is true if the iterate has changed.
[in]	iter	is the outer algorithm iterations count.

Reimplemented in ROL::InteriorPoint::PrimalDualSymmetrizer< Real >, ROL::InteriorPoint::PrimalDualSymmetrizer< Real >, ROL::PrimalDualInteriorPointBlock22< Real >, ROL::PrimalDualInteriorPointBlock21< Real >, ROL::NullSpaceOperator< Real >, ROL::StdLinearOperator< Real >, ROL::LinearOperatorSum< Real >, ROL::LinearOperatorProduct< Real >, ROL::PrimalDualInteriorPointBlock11< Real >, ROL::RangeSpaceOperator< Real >, ROL::BlockOperator2Determinant< Real >, and ROL::DiagonalOperator< Real >.

Definition at line 49 of file ROL_LinearOperator.hpp.

template<class Real>

virtual void ROL::LinearOperator< Real >::apply	(	Vector< Real > &	Hv,
		const Vector< Real > &	v,
		Real &	tol
	)		const

pure virtual

Apply linear operator.

This function applies the linear operator to a vector.

Parameters

[out]	Hv	is the output vector.
[in]	v	is the input vector.
[in]	tol	is a tolerance for inexact linear operator application.

Implemented in ROL::InteriorPoint::PrimalDualSymmetrizer< Real >, ROL::InteriorPoint::PrimalDualSymmetrizer< Real >, ROL::PrimalDualInteriorPointBlock22< Real >, ROL::BoundFletcher< Real >::AugSystemPrecond, ROL::BoundFletcher< Real >::AugSystemNonSym, ROL::PrimalDualActiveSetStep< Real >::PrecondPD, ROL::BoundFletcher< Real >::AugSystemSym, ROL::TypeB::PrimalDualActiveSetAlgorithm< Real >::PrecondPDAS_Poly, ROL::PrimalDualInteriorPointBlock21< Real >, ROL::Secant< Real >, ROL::PrimalDualActiveSetStep< Real >::HessianPD, ROL::TypeB::PrimalDualActiveSetAlgorithm< Real >::HessianPDAS_Poly, ROL::Fletcher< Real >::AugSystemPrecond, ROL::PrimalDualInteriorPointBlock12< Real >, ROL::ProjectedNewtonKrylovStep< Real >::PrecondPNK, ROL::TypeB::PrimalDualActiveSetAlgorithm< Real >::PrecondPDAS, ROL::TypeB::NewtonKrylovAlgorithm< Real >::PrecondPNK, ROL::Fletcher< Real >::AugSystem, ROL::FletcherObjectiveE< Real >::AugSystemPrecond, ROL::SemismoothNewtonProjection< Real >::Precond, IdentityOperator< Real >, ROL::NullSpaceOperator< Real >, ROL::SemismoothNewtonProjection< Real >::Jacobian, ROL::TypeB::PrimalDualActiveSetAlgorithm< Real >::HessianPDAS, ROL::FletcherObjectiveE< Real >::AugSystem, ROL::StdLinearOperator< Real >, ROL::TypeB::NewtonKrylovAlgorithm< Real >::HessianPNK, ROL::NewtonKrylovStep< Real >::PrecondNK, ROL::ProjectedNewtonKrylovStep< Real >::HessianPNK, ROL::HouseholderReflector< Real >, ROL::LinearOperatorSum< Real >, ROL::PrimalDualInteriorPointBlock11< Real >, ROL::NewtonKrylov_U< Real >::PrecondNK, ROL::NewtonKrylovStep< Real >::HessianNK, ROL::LinearOperatorProduct< Real >, TridiagonalToeplitzOperator< Real >, TridiagonalToeplitzOperator< Real >, ROL::BlockOperator2< Real >, ROL::BlockOperator2Determinant< Real >, ROL::NewtonKrylov_U< Real >::HessianNK, ROL::RangeSpaceOperator< Real >, ROL::DiagonalOperator< Real >, ROL::BlockOperator< Real >, ROL::LinearOperatorFromConstraint< Real >, ROL::AugmentedSystemOperator< Real >, ROL::DyadicOperator< Real >, ROL::AugmentedSystemPrecOperator< Real >, Identity< Real >, Identity< Real >, ROL::IdentityOperator< Real >, and ROL::NullOperator< Real >.

Referenced by ROL::LinearOperator< Real >::applyAdjoint(), ROL::Lanczos< Real >::iterate(), ROL::ConjugateGradients< Real >::run(), ROL::ConjugateResiduals< Real >::run(), ROL::BiCGSTAB< Real >::run(), ROL::GMRES< Real >::run(), and ROL::details::MINRES< Real >::run().

template<class Real>

virtual void ROL::LinearOperator< Real >::applyInverse	(	Vector< Real > &	Hv,
		const Vector< Real > &	v,
		Real &	tol
	)		const

inlinevirtual

Apply inverse of linear operator.

This function applies the inverse of linear operator to a vector.

Parameters

[out]	Hv	is the output vector.
[in]	v	is the input vector.
[in]	tol	is a tolerance for inexact linear operator application.

Definition at line 67 of file ROL_LinearOperator.hpp.

References ROL::Vector< Real >::set().

Referenced by ROL::LinearOperator< Real >::applyAdjointInverse(), ROL::ConjugateGradients< Real >::run(), ROL::ConjugateResiduals< Real >::run(), ROL::BiCGSTAB< Real >::run(), and ROL::GMRES< Real >::run().

template<class Real>

virtual void ROL::LinearOperator< Real >::applyAdjoint	(	Vector< Real > &	Hv,
		const Vector< Real > &	v,
		Real &	tol
	)		const

inlinevirtual

Apply adjoint of linear operator.

This function applies the adjoint of a linear operator to a vector. Default behavior assumes operator is self-adjoint.

Parameters

[out]	Hv	is the output vector.
[in]	v	is the input vector.
[in]	tol	is a tolerance for inexact linear operator application.

Reimplemented in ROL::NullSpaceOperator< Real >, ROL::StdLinearOperator< Real >, ROL::AugmentedSystemOperator< Real >, ROL::RangeSpaceOperator< Real >, and ROL::AugmentedSystemPrecOperator< Real >.

Definition at line 80 of file ROL_LinearOperator.hpp.

References ROL::LinearOperator< Real >::apply().

template<class Real>

virtual void ROL::LinearOperator< Real >::applyAdjointInverse	(	Vector< Real > &	Hv,
		const Vector< Real > &	v,
		Real &	tol
	)		const

inlinevirtual

Apply adjoint of the inverse linear operator.

This function applies the adjoint of the inverse linear operator to a vector. Default behavior assumes operator is self-adjoint.

Parameters

[out]	Hv	is the output vector.
[in]	v	is the input vector.
[in]	tol	is a tolerance for inexact linear operator application.

Reimplemented in ROL::StdLinearOperator< Real >, ROL::NullSpaceOperator< Real >, ROL::AugmentedSystemOperator< Real >, ROL::RangeSpaceOperator< Real >, and ROL::AugmentedSystemPrecOperator< Real >.

Definition at line 93 of file ROL_LinearOperator.hpp.

References ROL::LinearOperator< Real >::applyInverse().

The documentation for this class was generated from the following file:

ROL_LinearOperator.hpp

Public Member Functions

Detailed Description

template<class Real> class ROL::LinearOperator< Real >

Constructor & Destructor Documentation

Member Function Documentation

template<class Real>
class ROL::LinearOperator< Real >