Secant predictor strategy More...

#include <LOCA_MultiPredictor_Secant.H>

Inheritance diagram for LOCA::MultiPredictor::Secant:

Collaboration diagram for LOCA::MultiPredictor::Secant:

Public Member Functions
	Secant (const Teuchos::RCP< LOCA::GlobalData > &global_data, const Teuchos::RCP< LOCA::Parameter::SublistParser > &topParams, const Teuchos::RCP< Teuchos::ParameterList > &predParams)
	Constructor. More...

virtual	~Secant ()
	Destructor.

	Secant (const Secant &source, NOX::CopyType type=NOX::DeepCopy)
	Copy constructor.

virtual LOCA::MultiPredictor::AbstractStrategy &	operator= (const LOCA::MultiPredictor::AbstractStrategy &source)
	Assignment operator.

virtual Teuchos::RCP < LOCA::MultiPredictor::AbstractStrategy >	clone (NOX::CopyType type=NOX::DeepCopy) const
	Clone function.

virtual NOX::Abstract::Group::ReturnType	compute (bool baseOnSecant, const std::vector< double > &stepSize, LOCA::MultiContinuation::ExtendedGroup &grp, const LOCA::MultiContinuation::ExtendedVector &prevXVec, const LOCA::MultiContinuation::ExtendedVector &xVec)
	Compute the predictor given the current and previous solution vectors. Set baseOnSecant to false if the predictor orientation should not be based on the secant vector (first or last steps of a continuation run). More...

virtual NOX::Abstract::Group::ReturnType	evaluate (const std::vector< double > &stepSize, const LOCA::MultiContinuation::ExtendedVector &xVec, LOCA::MultiContinuation::ExtendedMultiVector &result) const
	Evaluate predictor with step size `stepSize`. More...

virtual NOX::Abstract::Group::ReturnType	computeTangent (LOCA::MultiContinuation::ExtendedMultiVector &tangent)
	Compute tangent to predictor and store in `tangent`.

virtual bool	isTangentScalable () const
	Is the tangent vector for this predictor scalable. More...

Public Member Functions inherited from LOCA::MultiPredictor::AbstractStrategy
	AbstractStrategy ()
	Constructor.

virtual	~AbstractStrategy ()
	Destructor.

Protected Attributes
Teuchos::RCP< LOCA::GlobalData >	globalData
	LOCA global data object.

Teuchos::RCP < LOCA::MultiPredictor::AbstractStrategy >	firstStepPredictor
	First step predictor.

bool	isFirstStep
	Flag indicating whether this is the first step.

bool	isFirstStepComputed
	Flag indicating whether first step predictor has been computed.

Teuchos::RCP < LOCA::MultiContinuation::ExtendedMultiVector >	predictor
	Stores predictor vector.

Teuchos::RCP < LOCA::MultiContinuation::ExtendedVector >	secant
	Stores secant vector for setting orientation.

bool	initialized
	Flag indicating whether vectors have been initialized.

Additional Inherited Members
Protected Member Functions inherited from LOCA::MultiPredictor::AbstractStrategy
virtual void	setPredictorOrientation (bool baseOnSecant, const std::vector< double > &stepSize, const LOCA::MultiContinuation::ExtendedGroup &grp, const LOCA::MultiContinuation::ExtendedVector &prevXVec, const LOCA::MultiContinuation::ExtendedVector &xVec, LOCA::MultiContinuation::ExtendedVector &secant, LOCA::MultiContinuation::ExtendedMultiVector &tangent)
	Sets orientation of predictor based on parameter change from previous steps. More...

Detailed Description

Secant predictor strategy

This class implements a predictor strategy based on computing the secant vector $v$ to the continuation curve given by

$v = x - x_o$

where $x$ is the current solution vector and $x_o$ is the previous solution vector. Note that for multi-parameter continuation, the solution component for each secant direction is given as above, with the parameter components given by the identity matrix.

For the first step of a continuation run, $x_o$ is not defined, and so a different predictor is used for this step. This predictor is specified via the "First Step Predictor" sublist of the "Predictor" sublist. This predictor is instantiated using the LOCA::Factory as usual.

The parameters used by this class supplied in the constructor are:

"First Step Predictor" Predictor sublist for first step

Constructor & Destructor Documentation

LOCA::MultiPredictor::Secant::Secant	(	const Teuchos::RCP< LOCA::GlobalData > &	global_data,
		const Teuchos::RCP< LOCA::Parameter::SublistParser > &	topParams,
		const Teuchos::RCP< Teuchos::ParameterList > &	predParams
	)

Constructor.

Parameters

global_data	[in] Global data object used for LOCA::Factory
topParams	[in] Parsed top-level parameter list used when creating first step predictor
predParams	[in] Predictor parameters used to obtain "First Step Predictor" as described above.

References firstStepPredictor, Teuchos::ParameterList::get(), LOCA::Parameter::SublistParser::getSublist(), globalData, and LOCA::GlobalData::locaFactory.

Member Function Documentation

NOX::Abstract::Group::ReturnType LOCA::MultiPredictor::Secant::compute	(	bool	baseOnSecant,
		const std::vector< double > &	stepSize,
		LOCA::MultiContinuation::ExtendedGroup &	grp,
		const LOCA::MultiContinuation::ExtendedVector &	prevXVec,
		const LOCA::MultiContinuation::ExtendedVector &	xVec
	)

virtual

Compute the predictor given the current and previous solution vectors. Set baseOnSecant to false if the predictor orientation should not be based on the secant vector (first or last steps of a continuation run).

This method actually implements the secant calculation described above

Implements LOCA::MultiPredictor::AbstractStrategy.

References LOCA::MultiContinuation::ExtendedVector::clone(), LOCA::Extended::Vector::createMultiVector(), NOX::Abstract::Group::Ok, NOX::Abstract::Vector::scale(), NOX::ShapeCopy, NOX::Utils::StepperDetails, and LOCA::Extended::Vector::update().

NOX::Abstract::Group::ReturnType LOCA::MultiPredictor::Secant::evaluate	(	const std::vector< double > &	stepSize,
		const LOCA::MultiContinuation::ExtendedVector &	xVec,
		LOCA::MultiContinuation::ExtendedMultiVector &	result
	)		const

virtual

Evaluate predictor with step size stepSize.

This method computes result[i] = xVec[i] + stepSize[i] * v[i] for each i, where v[i] is the ith predictor direction.

Implements LOCA::MultiPredictor::AbstractStrategy.

References NOX::Abstract::Group::Ok.

bool LOCA::MultiPredictor::Secant::isTangentScalable ( ) const

virtual

Is the tangent vector for this predictor scalable.

For the secant predictor, this always returns true.

Implements LOCA::MultiPredictor::AbstractStrategy.

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

LOCA_MultiPredictor_Secant.H
LOCA_MultiPredictor_Secant.C

Public Member Functions

Protected Attributes

Additional Inherited Members

Detailed Description

Constructor & Destructor Documentation

Member Function Documentation