ModelEvaluator for forming adjoint sensitivity equations. More...

#include <Tempus_AdjointSensitivityModelEvaluator_decl.hpp>

Inheritance diagram for Tempus::AdjointSensitivityModelEvaluator< Scalar >:

Public Types
typedef Thyra::VectorBase< Scalar >	Vector

typedef Thyra::MultiVectorBase < Scalar >	MultiVector

Public Types inherited from Thyra::ModelEvaluator< class >
enum	EInArgsMembers

enum	EInArgs_p_mp

enum	EEvalType

enum	EDerivativeMultiVectorOrientation

enum	EDerivativeLinearOp

enum	EDerivativeLinearity

enum	ERankStatus

enum	EOutArgsMembers

enum	EOutArgsDfDp

enum	EOutArgsDgDx_dot

enum	EOutArgsDgDx

enum	EOutArgsDgDp

enum	EOutArgsDfDp_mp

enum	EOutArgs_g_mp

enum	EOutArgsDgDx_dot_mp

enum	EOutArgsDgDx_mp

enum	EOutArgsDgDp_mp

typedef Teuchos::ScalarTraits < Scalar >::magnitudeType	ScalarMag

Public Member Functions
	AdjointSensitivityModelEvaluator (const Teuchos::RCP< const Thyra::ModelEvaluator< Scalar > > &model, const Teuchos::RCP< const Thyra::ModelEvaluator< Scalar > > &adjoint_residual_model, const Teuchos::RCP< const Thyra::ModelEvaluator< Scalar > > &adjoint_solve_model, const Scalar &t_init, const Scalar &t_final, const bool is_pseudotransient, const Teuchos::RCP< const Teuchos::ParameterList > &pList=Teuchos::null)
	Constructor. More...

Teuchos::RCP< const Thyra::ModelEvaluator< Scalar > >	getModel () const
	Get the underlying model 'f'. More...

Teuchos::RCP< const Thyra::ModelEvaluator< Scalar > >	getAdjointResidualModel () const
	Get the underlying adjoint residual model. More...

Teuchos::RCP< const Thyra::ModelEvaluator< Scalar > >	getAdjointSolveModel () const
	Get the underlying adjoint solve model. More...

void	setFinalTime (const Scalar t_final)
	Set the final time from the forward evaluation. More...

void	setForwardSolutionHistory (const Teuchos::RCP< const Tempus::SolutionHistory< Scalar > > &sh)
	Set solution history from forward evaluation. More...

Public Member Functions inherited from Thyra::StateFuncModelEvaluatorBase< Scalar >
RCP< const VectorSpaceBase < Scalar > >	get_p_space (int l) const

RCP< const Teuchos::Array < std::string > >	get_p_names (int l) const

RCP< const VectorSpaceBase < Scalar > >	get_g_space (int j) const

Teuchos::ArrayView< const std::string >	get_g_names (int j) const

ModelEvaluatorBase::InArgs < Scalar >	getNominalValues () const

ModelEvaluatorBase::InArgs < Scalar >	getLowerBounds () const

ModelEvaluatorBase::InArgs < Scalar >	getUpperBounds () const

RCP< LinearOpBase< Scalar > >	create_W_op () const

RCP< PreconditionerBase< Scalar > >	create_W_prec () const

RCP< const LinearOpWithSolveFactoryBase < Scalar > >	get_W_factory () const

void	reportFinalPoint (const ModelEvaluatorBase::InArgs< Scalar > &finalPoint, const bool wasSolved)

Public Member Functions inherited from Thyra::ModelEvaluator< class >
virtual int	Np () const =0

virtual int	Ng () const =0

virtual RCP< const VectorSpaceBase< Scalar > >	get_f_multiplier_space () const =0

virtual RCP< const VectorSpaceBase< Scalar > >	get_g_multiplier_space (int j) const =0

virtual RCP < LinearOpWithSolveBase < Scalar > >	create_W () const =0

virtual RCP< LinearOpBase < Scalar > >	create_DfDp_op (int l) const =0

virtual RCP< LinearOpBase < Scalar > >	create_DgDx_dot_op (int j) const =0

virtual RCP< LinearOpBase < Scalar > >	create_DgDx_op (int j) const =0

virtual RCP< LinearOpBase < Scalar > >	create_DgDp_op (int j, int l) const =0

virtual ModelEvaluatorBase::OutArgs < Scalar >	createOutArgs () const =0

virtual void	evalModel (const ModelEvaluatorBase::InArgs< Scalar > &inArgs, const ModelEvaluatorBase::OutArgs< Scalar > &outArgs) const =0

	ModelEvaluatorBase ()

std::string	toString (ModelEvaluatorBase::EInArgsMembers)

std::string	toString (ModelEvaluatorBase::EOutArgsMembers)

std::string	toString (ModelEvaluatorBase::EDerivativeMultiVectorOrientation orientation)

ModelEvaluatorBase::EDerivativeMultiVectorOrientation	getOtherDerivativeMultiVectorOrientation (ModelEvaluatorBase::EDerivativeMultiVectorOrientation orientation)

Public Member Functions inherited from Teuchos::Describable
virtual std::string	description () const

virtual void	describe (FancyOStream &out, const EVerbosityLevel verbLevel=verbLevel_default) const

void	describe (std::ostream &out, const EVerbosityLevel verbLevel=verbLevel_default) const

virtual	~Describable ()

	LabeledObject ()

virtual	~LabeledObject ()

virtual void	setObjectLabel (const std::string &objectLabel)

virtual std::string	getObjectLabel () const

DescribableStreamManipulatorState	describe (const Describable &describable, const EVerbosityLevel verbLevel=Describable::verbLevel_default)

std::ostream &	operator<< (std::ostream &os, const DescribableStreamManipulatorState &d)

Static Public Member Functions
static Teuchos::RCP< const Teuchos::ParameterList >	getValidParameters ()

Private Types
typedef Thyra::DefaultMultiVectorProductVectorSpace < Scalar >	DMVPVS

typedef Thyra::DefaultMultiVectorProductVector < Scalar >	DMVPV

Private Member Functions
Thyra::ModelEvaluatorBase::OutArgs < Scalar >	createOutArgsImpl () const

void	evalModelImpl (const Thyra::ModelEvaluatorBase::InArgs< Scalar > &inArgs, const Thyra::ModelEvaluatorBase::OutArgs< Scalar > &outArgs) const

Private Attributes
Thyra::ModelEvaluatorBase::InArgs < Scalar >	prototypeInArgs_

Thyra::ModelEvaluatorBase::OutArgs < Scalar >	prototypeOutArgs_

Teuchos::RCP< const Thyra::ModelEvaluator< Scalar > >	model_

Teuchos::RCP< const Thyra::ModelEvaluator< Scalar > >	adjoint_residual_model_

Teuchos::RCP< const Thyra::ModelEvaluator< Scalar > >	adjoint_solve_model_

Teuchos::RCP< const DMVPVS >	adjoint_space_

Teuchos::RCP< const DMVPVS >	residual_space_

Teuchos::RCP< const DMVPVS >	response_space_

Teuchos::RCP< const Tempus::SolutionHistory < Scalar > >	sh_

Scalar	t_init_

Scalar	t_final_

bool	is_pseudotransient_

bool	mass_matrix_is_constant_

bool	mass_matrix_is_identity_

int	p_index_

int	g_index_

int	num_adjoint_

bool	mass_matrix_is_computed_

bool	jacobian_matrix_is_computed_

bool	response_gradient_is_computed_

Teuchos::RCP < Thyra::VectorBase< Scalar > >	my_x_dot_

Teuchos::RCP < Thyra::LinearOpBase< Scalar > >	my_dfdx_

Teuchos::RCP < Thyra::LinearOpBase< Scalar > >	my_dfdxdot_

Teuchos::RCP < Thyra::LinearOpBase< Scalar > >	my_dfdp_op_

Teuchos::RCP < Thyra::MultiVectorBase < Scalar > >	my_dfdp_mv_

Teuchos::RCP < Thyra::MultiVectorBase < Scalar > >	my_dgdx_mv_

Teuchos::RCP < Tempus::SolutionState < Scalar > >	forward_state_

Scalar	t_interp_

Public functions overridden from ModelEvaulator.
Teuchos::RCP< const Thyra::VectorSpaceBase< Scalar > >	get_p_space (int p) const

Teuchos::RCP< const Teuchos::Array< std::string > >	get_p_names (int p) const

Teuchos::RCP< const Thyra::VectorSpaceBase< Scalar > >	get_x_space () const

Teuchos::RCP< const Thyra::VectorSpaceBase< Scalar > >	get_f_space () const

Teuchos::RCP< const Thyra::VectorSpaceBase< Scalar > >	get_g_space (int j) const

Teuchos::RCP < Thyra::LinearOpBase< Scalar > >	create_W_op () const

Teuchos::RCP< const Thyra::LinearOpWithSolveFactoryBase < Scalar > >	get_W_factory () const

Thyra::ModelEvaluatorBase::InArgs < Scalar >	createInArgs () const

Thyra::ModelEvaluatorBase::InArgs < Scalar >	getNominalValues () const

Additional Inherited Members
Static Public Attributes inherited from Thyra::ModelEvaluator< class >
static const int	NUM_E_IN_ARGS_MEMBERS

static const int	NUM_E_OUT_ARGS_MEMBERS

Static Public Attributes inherited from Teuchos::Describable
static const EVerbosityLevel	verbLevel_default

Detailed Description

template<typename Scalar>
class Tempus::AdjointSensitivityModelEvaluator< Scalar >

ModelEvaluator for forming adjoint sensitivity equations.

This class wraps a given ModelEvalutor encapsulating f(x_dot,x,p) and creates a new "residual" for the adjoint sensitivity equations: F(y) = d/dt( df/dx_dot^T*y ) - df/dx^T*y + dg/dx^T = 0 where y is the adjoint variable. Nominally y is a multi-vector belonging to the vector space of the residual f with number of columns equal to the dimension of g. To satisfy the model evaluator interface, y is converted to a product vector formed by its columns. This is the form of the adjoint equations suitable for computing pseudotransientsensitivities of a response function g(x(T),p) or adjoint sensitivities for a time-integrated response function int_0^T g(x(t),p).

To compute adjoint sensitivities, the equations f(x_dot,x,p) = 0 are integrated forward in time to some final time T, with the adjoint equations above integrated backward in time starting at T. Since the tempus integrator can only integrate forward in time, we define tau = T - t and transform the adjoint equations to: F(y) = d/dtau( df/dx_dot^T*y ) + df/dx^T*y - dg/dx^T = 0. The initial conditions for y are y(T) = 0. The sensitivity of g(T) is then int_0^T(dg/dp(T) - df/dp^T*y)dt + dx/dp(0)^T*df/dx_dot(0)^T*y(0) for transient adjoint sensitivites and is dg/dp(T) - df/dp^T*y for pseudotransient.

This model evaluator supports both implicit and explict forms of the governing equations. However it assumes df/dx_dot is constant with respect to x_dot, x, and t so that the adjoint equations become F(y) = df/dxdot^T*y_dot + df/dx^T*y - dg/dx^T = 0. Moving beyond this assumption will require modifications to the steppers in how they generate time-derivative terms.

Definition at line 55 of file Tempus_AdjointSensitivityModelEvaluator_decl.hpp.

Member Typedef Documentation

template<typename Scalar >

typedef Thyra::VectorBase<Scalar> Tempus::AdjointSensitivityModelEvaluator< Scalar >::Vector

Definition at line 58 of file Tempus_AdjointSensitivityModelEvaluator_decl.hpp.

template<typename Scalar >

typedef Thyra::MultiVectorBase<Scalar> Tempus::AdjointSensitivityModelEvaluator< Scalar >::MultiVector

Definition at line 59 of file Tempus_AdjointSensitivityModelEvaluator_decl.hpp.

template<typename Scalar >

typedef Thyra::DefaultMultiVectorProductVectorSpace<Scalar> Tempus::AdjointSensitivityModelEvaluator< Scalar >::DMVPVS

private

Definition at line 144 of file Tempus_AdjointSensitivityModelEvaluator_decl.hpp.

template<typename Scalar >

typedef Thyra::DefaultMultiVectorProductVector<Scalar> Tempus::AdjointSensitivityModelEvaluator< Scalar >::DMVPV

private

Definition at line 145 of file Tempus_AdjointSensitivityModelEvaluator_decl.hpp.

Constructor & Destructor Documentation

template<typename Scalar >

Tempus::AdjointSensitivityModelEvaluator< Scalar >::AdjointSensitivityModelEvaluator	(	const Teuchos::RCP< const Thyra::ModelEvaluator< Scalar > > &	model,
		const Teuchos::RCP< const Thyra::ModelEvaluator< Scalar > > &	adjoint_residual_model,
		const Teuchos::RCP< const Thyra::ModelEvaluator< Scalar > > &	adjoint_solve_model,
		const Scalar &	t_init,
		const Scalar &	t_final,
		const bool	is_pseudotransient,
		const Teuchos::RCP< const Teuchos::ParameterList > &	pList = `Teuchos::null`
	)

Constructor.

t_final is the final integration time used to implement the time transformation described above. num_adjoint is the number of adjoint variables defining the number of columns in y, which is determined by the number of elements in the vector space for the response g. The optionally supplied parameter list supports the following options:

"Sensitivity Parameter Index", (default: 0) The model evaluator parameter index for which sensitivities will be computed.
"Response Function Index", (default: 0) The model evaluator response index for which sensitivities will be computed.
"Mass Matrix Is Constant" (default: true) Whether the mass matrix df/dx_dot is a constant matrix. As describe above, this is currently required to be true.
"Mass Matrix Is Identity" (default: false) Whether the mass matrix is the identity matrix, in which some computations can be skipped.

Definition at line 29 of file Tempus_AdjointSensitivityModelEvaluator_impl.hpp.

Member Function Documentation

template<typename Scalar >

Teuchos::RCP<const Thyra::ModelEvaluator<Scalar> > Tempus::AdjointSensitivityModelEvaluator< Scalar >::getModel ( ) const

inline

Get the underlying model 'f'.

Definition at line 91 of file Tempus_AdjointSensitivityModelEvaluator_decl.hpp.

template<typename Scalar >

Teuchos::RCP<const Thyra::ModelEvaluator<Scalar> > Tempus::AdjointSensitivityModelEvaluator< Scalar >::getAdjointResidualModel ( ) const

inline

Get the underlying adjoint residual model.

Definition at line 97 of file Tempus_AdjointSensitivityModelEvaluator_decl.hpp.

template<typename Scalar >

Teuchos::RCP<const Thyra::ModelEvaluator<Scalar> > Tempus::AdjointSensitivityModelEvaluator< Scalar >::getAdjointSolveModel ( ) const

inline

Get the underlying adjoint solve model.

Definition at line 104 of file Tempus_AdjointSensitivityModelEvaluator_decl.hpp.

template<typename Scalar >

void Tempus::AdjointSensitivityModelEvaluator< Scalar >::setFinalTime ( const Scalar t_final )

Set the final time from the forward evaluation.

Definition at line 123 of file Tempus_AdjointSensitivityModelEvaluator_impl.hpp.

template<typename Scalar >

void Tempus::AdjointSensitivityModelEvaluator< Scalar >::setForwardSolutionHistory ( const Teuchos::RCP< const Tempus::SolutionHistory< Scalar > > & sh )

Set solution history from forward evaluation.

Definition at line 130 of file Tempus_AdjointSensitivityModelEvaluator_impl.hpp.

template<typename Scalar >

Teuchos::RCP< const Thyra::VectorSpaceBase< Scalar > > Tempus::AdjointSensitivityModelEvaluator< Scalar >::get_p_space ( int p ) const

virtual

Implements Thyra::ModelEvaluator< class >.

Definition at line 149 of file Tempus_AdjointSensitivityModelEvaluator_impl.hpp.

template<typename Scalar >

Teuchos::RCP< const Teuchos::Array< std::string > > Tempus::AdjointSensitivityModelEvaluator< Scalar >::get_p_names ( int p ) const

virtual

Implements Thyra::ModelEvaluator< class >.

Definition at line 157 of file Tempus_AdjointSensitivityModelEvaluator_impl.hpp.

template<typename Scalar >

Teuchos::RCP< const Thyra::VectorSpaceBase< Scalar > > Tempus::AdjointSensitivityModelEvaluator< Scalar >::get_x_space ( ) const

virtual

Implements Thyra::ModelEvaluator< class >.

Definition at line 165 of file Tempus_AdjointSensitivityModelEvaluator_impl.hpp.

template<typename Scalar >

Teuchos::RCP< const Thyra::VectorSpaceBase< Scalar > > Tempus::AdjointSensitivityModelEvaluator< Scalar >::get_f_space ( ) const

virtual

Implements Thyra::ModelEvaluator< class >.

Definition at line 172 of file Tempus_AdjointSensitivityModelEvaluator_impl.hpp.

template<typename Scalar >

Teuchos::RCP< const Thyra::VectorSpaceBase< Scalar > > Tempus::AdjointSensitivityModelEvaluator< Scalar >::get_g_space ( int j ) const

virtual

Implements Thyra::ModelEvaluator< class >.

Definition at line 179 of file Tempus_AdjointSensitivityModelEvaluator_impl.hpp.

template<typename Scalar >

Teuchos::RCP< Thyra::LinearOpBase< Scalar > > Tempus::AdjointSensitivityModelEvaluator< Scalar >::create_W_op ( ) const

virtual

Implements Thyra::ModelEvaluator< class >.

Definition at line 188 of file Tempus_AdjointSensitivityModelEvaluator_impl.hpp.

template<typename Scalar >

Teuchos::RCP< const Thyra::LinearOpWithSolveFactoryBase< Scalar > > Tempus::AdjointSensitivityModelEvaluator< Scalar >::get_W_factory ( ) const

virtual

Implements Thyra::ModelEvaluator< class >.

Definition at line 199 of file Tempus_AdjointSensitivityModelEvaluator_impl.hpp.

template<typename Scalar >

Thyra::ModelEvaluatorBase::InArgs< Scalar > Tempus::AdjointSensitivityModelEvaluator< Scalar >::createInArgs ( ) const

virtual

Implements Thyra::ModelEvaluator< class >.

Definition at line 215 of file Tempus_AdjointSensitivityModelEvaluator_impl.hpp.

template<typename Scalar >

Thyra::ModelEvaluatorBase::InArgs< Scalar > Tempus::AdjointSensitivityModelEvaluator< Scalar >::getNominalValues ( ) const

virtual

Implements Thyra::ModelEvaluator< class >.

Definition at line 222 of file Tempus_AdjointSensitivityModelEvaluator_impl.hpp.

template<class Scalar >

Teuchos::RCP< const Teuchos::ParameterList > Tempus::AdjointSensitivityModelEvaluator< Scalar >::getValidParameters ( )

static

Definition at line 541 of file Tempus_AdjointSensitivityModelEvaluator_impl.hpp.

template<typename Scalar >

Thyra::ModelEvaluatorBase::OutArgs< Scalar > Tempus::AdjointSensitivityModelEvaluator< Scalar >::createOutArgsImpl ( ) const

private

Definition at line 253 of file Tempus_AdjointSensitivityModelEvaluator_impl.hpp.

template<typename Scalar >

void Tempus::AdjointSensitivityModelEvaluator< Scalar >::evalModelImpl	(	const Thyra::ModelEvaluatorBase::InArgs< Scalar > &	inArgs,
		const Thyra::ModelEvaluatorBase::OutArgs< Scalar > &	outArgs
	)		const