NOX Development
Public Member Functions | Protected Member Functions | Protected Attributes
NOX::Thyra::Group Class Reference

A concrete implementation of the NOX::Abstract::Group using Thyra. More...

#include <NOX_Thyra_Group.H>

Inheritance diagram for NOX::Thyra::Group:
Inheritance graph
[legend]
Collaboration diagram for NOX::Thyra::Group:
Collaboration graph
[legend]

List of all members.

Public Member Functions

 Group (const NOX::Thyra::Vector &initial_guess, const Teuchos::RCP< const ::Thyra::ModelEvaluator< double > > &model, const Teuchos::RCP< const ::Thyra::VectorBase< double > > &weight_vector=Teuchos::null)
 The default constructor that uses the linear solver from the ModelEvaluator.
 Group (const NOX::Thyra::Vector &initial_guess, const Teuchos::RCP< const ::Thyra::ModelEvaluator< double > > &model, const Teuchos::RCP< ::Thyra::LinearOpBase< double > > &linear_op, const Teuchos::RCP< const ::Thyra::LinearOpWithSolveFactoryBase< double > > &lows_factory, const Teuchos::RCP< ::Thyra::PreconditionerBase< double > > &prec_op, const Teuchos::RCP< ::Thyra::PreconditionerFactoryBase< double > > &prec_factory, const Teuchos::RCP< const ::Thyra::VectorBase< double > > &weight_vector=Teuchos::null)
 Power user constructor that takes explicit linear solver objects to handle different combinations.
 Group (const NOX::Thyra::Group &source, NOX::CopyType type=DeepCopy)
 Copy constructor.
 ~Group ()
 Destructor.
NOX::Abstract::Groupoperator= (const NOX::Abstract::Group &source)
 Copies the source group into this group.
NOX::Abstract::Groupoperator= (const NOX::Thyra::Group &source)
Teuchos::RCP< const
::Thyra::VectorBase< double > > 
get_current_x () const
Teuchos::RCP
< ::Thyra::LinearOpBase
< double > > 
getNonconstJacobianOperator ()
Teuchos::RCP< const
::Thyra::LinearOpBase< double > > 
getJacobianOperator () const
Teuchos::RCP
< ::Thyra::LinearOpWithSolveBase
< double > > 
getNonconstJacobian ()
Teuchos::RCP< const
::Thyra::LinearOpWithSolveBase
< double > > 
getJacobian () const
virtual Teuchos::RCP
< NOX::Abstract::Group
clone (NOX::CopyType type=NOX::DeepCopy) const
 Create a new Group of the same derived type as this one by cloning this one, and return a ref count pointer to the new group.
void print () const
 Print out the group.
::Thyra::ModelEvaluatorBase::InArgs
< double > & 
getNonconstInArgs ()
const
::Thyra::ModelEvaluatorBase::InArgs
< double > & 
getInArgs () const
Teuchos::RCP< const
::Thyra::ModelEvaluator
< double > > 
getModel () const
"Compute" functions.
void setX (const NOX::Abstract::Vector &y)
 Set the solution vector x to y.
void setX (const NOX::Thyra::Vector &y)
 See above.
void computeX (const NOX::Abstract::Group &grp, const NOX::Abstract::Vector &d, double step)
 Compute x = grp.x + step * d.
void computeX (const NOX::Thyra::Group &grp, const NOX::Thyra::Vector &d, double step)
 See above.
NOX::Abstract::Group::ReturnType computeF ()
 Compute and store F(x).
NOX::Abstract::Group::ReturnType computeJacobian ()
 Compute and store Jacobian.
NOX::Abstract::Group::ReturnType computeGradient ()
 Compute and store gradient.
NOX::Abstract::Group::ReturnType computeNewton (Teuchos::ParameterList &params)
 Compute the Newton direction, using parameters for the linear solve.
Jacobian operations.

Operations using the Jacobian matrix. These may not be defined in matrix-free scenarios.

NOX::Abstract::Group::ReturnType applyJacobian (const NOX::Abstract::Vector &input, NOX::Abstract::Vector &result) const
 Applies Jacobian to the given input vector and puts the answer in the result.
NOX::Abstract::Group::ReturnType applyJacobian (const NOX::Thyra::Vector &input, NOX::Thyra::Vector &result) const
NOX::Abstract::Group::ReturnType applyJacobianMultiVector (const NOX::Abstract::MultiVector &input, NOX::Abstract::MultiVector &result) const
 applyJacobian for multiple right-hand sides
NOX::Abstract::Group::ReturnType applyJacobianTranspose (const NOX::Abstract::Vector &input, NOX::Abstract::Vector &result) const
 Applies Jacobian-Transpose to the given input vector and puts the answer in the result.
NOX::Abstract::Group::ReturnType applyJacobianTranspose (const NOX::Thyra::Vector &input, NOX::Thyra::Vector &result) const
NOX::Abstract::Group::ReturnType applyJacobianTransposeMultiVector (const NOX::Abstract::MultiVector &input, NOX::Abstract::MultiVector &result) const
 applyJacobianTranspose for multiple right-hand sides
NOX::Abstract::Group::ReturnType applyJacobianInverse (Teuchos::ParameterList &params, const NOX::Abstract::Vector &input, NOX::Abstract::Vector &result) const
 Applies the inverse of the Jacobian matrix to the given input vector and puts the answer in result.
NOX::Abstract::Group::ReturnType applyJacobianInverse (Teuchos::ParameterList &params, const NOX::Thyra::Vector &input, NOX::Thyra::Vector &result) const
NOX::Abstract::Group::ReturnType applyJacobianInverseMultiVector (Teuchos::ParameterList &params, const NOX::Abstract::MultiVector &input, NOX::Abstract::MultiVector &result) const
 applyJacobianInverse for multiple right-hand sides
NOX::Abstract::Group::ReturnType applyRightPreconditioning (bool useTranspose, Teuchos::ParameterList &params, const NOX::Abstract::Vector &input, NOX::Abstract::Vector &result) const
 Apply right preconditiong to the given input vector.
"Is" functions

Checks to see if various objects have been computed. Returns true if the corresponding "compute" function has been called since the last update to the solution vector (via instantiation or computeX).

bool isF () const
 Return true if F is valid.
bool isJacobian () const
 Return true if the Jacobian is valid.
bool isGradient () const
 Return true if the gradient is valid.
bool isNewton () const
 Return true if the Newton direction is valid.
"Get" functions

Note that these function do not check whether or not the vectors are valid. Must use the "Is" functions for that purpose.

const NOX::Abstract::VectorgetX () const
 Return solution vector.
const NOX::Abstract::VectorgetF () const
 Return F(x)
double getNormF () const
 Return 2-norm of F(x).
const NOX::Abstract::VectorgetGradient () const
 Return gradient.
const NOX::Abstract::VectorgetNewton () const
 Return Newton direction.
Teuchos::RCP< const
NOX::Abstract::Vector
getXPtr () const
 Return RCP to solution vector.
Teuchos::RCP< const
NOX::Abstract::Vector
getFPtr () const
 Return RCP to F(x)
Teuchos::RCP< const
NOX::Abstract::Vector
getGradientPtr () const
 Return RCP to gradient.
Teuchos::RCP< const
NOX::Abstract::Vector
getNewtonPtr () const
 Return RCP to Newton direction.

Protected Member Functions

void resetIsValidFlags ()
 resets the isValid flags to false
NOX::Abstract::Group::ReturnType applyJacobianInverseMultiVector (Teuchos::ParameterList &p, const ::Thyra::MultiVectorBase< double > &input,::Thyra::MultiVectorBase< double > &result) const
 Apply Jacobian inverse using Thyra objects.
::Thyra::ESolveMeasureNormType getThyraNormType (const std::string &name) const
void updateLOWS () const
 Finalizes LOWS to be a valid solver for the Jacobian.
void scaleResidualAndJacobian () const
void unscaleResidualAndJacobian () const

Protected Attributes

Teuchos::RCP< const
::Thyra::ModelEvaluator
< double > > 
model_
 Problem interface.
Teuchos::RCP< NOX::Thyra::Vectorx_vec_
 Solution vector.
Teuchos::RCP< NOX::Thyra::Vectorf_vec_
 Residual vector.
Teuchos::RCP< NOX::Thyra::Vectornewton_vec_
 Newton direction vector.
Teuchos::RCP< NOX::Thyra::Vectorgradient_vec_
 Gradient direction vector.
Teuchos::RCP
< NOX::SharedObject
< ::Thyra::LinearOpWithSolveBase
< double >, NOX::Thyra::Group > > 
shared_jacobian_
 Shared Jacobian operator with solve.
Teuchos::RCP
< ::Thyra::LinearOpBase
< double > > 
lop_
 Jacobian operator.
Teuchos::RCP< const
::Thyra::LinearOpWithSolveFactoryBase
< double > > 
lows_factory_
 Thyra LOWS factory for building Jacobians.
Teuchos::RCP< const
::Thyra::DefaultLinearOpSource
< double > > 
losb_
 Source base needed to create preconditioner.
Teuchos::RCP
< ::Thyra::PreconditionerBase
< double > > 
prec_
 Preconditioner for Jacobian.
Teuchos::RCP
< ::Thyra::PreconditionerFactoryBase
< double > > 
prec_factory_
 Preconditioner factory.
mutable::Thyra::ModelEvaluatorBase::InArgs
< double > 
in_args_
 Residual InArgs.
mutable::Thyra::ModelEvaluatorBase::OutArgs
< double > 
out_args_
 Residual OutArgs.
Teuchos::RCP< const
::Thyra::VectorBase< double > > 
weight_vec_
 Optional wieghting vector for function scaling.
Teuchos::RCP
< ::Thyra::VectorBase< double > > 
inv_weight_vec_
 Inverse of weight vector used to unscale problem.
IsValid flags

True if the current solution is up-to-date with respect to the currect solution vector.

bool is_valid_f_
bool is_valid_jacobian_
bool is_valid_newton_dir_
bool is_valid_gradient_dir_
bool is_valid_lows_

Detailed Description

A concrete implementation of the NOX::Abstract::Group using Thyra.

NOTE: This Group supports row sum scaling of the function (residual and Jacobian). This is enabled by setting a weight vector on the initial guess vector in the Group constructor. The residual and Jacobian must be scaled before and then unscaled after calls to construct the preconditioner and solve the linear system. This follows the nox epetra group. This design should be changed in a future nox refactor, but requires significant changes to the Group object.


Constructor & Destructor Documentation

NOX::Thyra::Group::Group ( const NOX::Thyra::Vector initial_guess,
const Teuchos::RCP< const ::Thyra::ModelEvaluator< double > > &  model,
const Teuchos::RCP< const ::Thyra::VectorBase< double > > &  weight_vector = Teuchos::null 
)

The default constructor that uses the linear solver from the ModelEvaluator.

Most users should use this constructor. It is meant to be used in conjunction with a stratimikos linear solver that is built as part of the input model evaluator. For finer grained control over the use of the preconditioner and for Jacobian-Free Newton-Krylov cases, the power user constructor should be used.

Parameters:
[in]initial_guessInitial guess for the solution vector
[in]modelModelEvaluator
[in]weight_vectorOptional diagonal weighting vector for the model.

References NOX::DeepCopy, f_vec_, gradient_vec_, in_args_, lop_, losb_, lows_factory_, model_, newton_vec_, nonnull(), out_args_, prec_, prec_factory_, Teuchos::rcp(), resetIsValidFlags(), NOX::ShapeCopy, shared_jacobian_, weight_vec_, and x_vec_.

NOX::Thyra::Group::Group ( const NOX::Thyra::Vector initial_guess,
const Teuchos::RCP< const ::Thyra::ModelEvaluator< double > > &  model,
const Teuchos::RCP< ::Thyra::LinearOpBase< double > > &  linear_op,
const Teuchos::RCP< const ::Thyra::LinearOpWithSolveFactoryBase< double > > &  lows_factory,
const Teuchos::RCP< ::Thyra::PreconditionerBase< double > > &  prec_op,
const Teuchos::RCP< ::Thyra::PreconditionerFactoryBase< double > > &  prec_factory,
const Teuchos::RCP< const ::Thyra::VectorBase< double > > &  weight_vector = Teuchos::null 
)

Power user constructor that takes explicit linear solver objects to handle different combinations.

This class allows the user to set user-defined linear operators and preconditioners (and corresponding factories). The user can set the linear_op to be a Jacobian-Free Newton Krylov operator (use the class NOX::Thyra::MatrixFreeJacobianOperator).

Parameters:
[in]initial_guess(Required) Initial guess for the solution vector
[in]model(Required) ModelEvaluator
[in]linear_op(Optional) Forward operator for the Jacobian. Must be non-null for Newton-based solvers.
[in]lows_factory(Optional) Factory for building and updating linear solver.
[in]prec_op(Optional) Preconditioner operator. If set to Teuchos::null and a non-null prec_factory exists, the prec_op will be constructed using the preconditioner factory.
[in]prec_factory(Optional) Factory for updating the precondiitoner. If set to Teuchos::null and there is a non-null prec_op, then the preconditioner will be updated using the model evaluator as long as the ModelEvaluator::outArgs supports W_prec.
[in]weight_vector(Optional) diagonal weighting vector for the model.

References NOX::DeepCopy, f_vec_, gradient_vec_, in_args_, is_null(), lop_, losb_, lows_factory_, model_, newton_vec_, nonnull(), out_args_, prec_, prec_factory_, Teuchos::rcp(), resetIsValidFlags(), NOX::ShapeCopy, shared_jacobian_, weight_vec_, and x_vec_.


Member Function Documentation

NOX::Abstract::Group::ReturnType NOX::Thyra::Group::applyJacobian ( const NOX::Abstract::Vector input,
NOX::Abstract::Vector result 
) const [virtual]

Applies Jacobian to the given input vector and puts the answer in the result.

Computes

\[ v = J u, \]

where $J$ is the Jacobian, $u$ is the input vector, and $v$ is the result vector.

Returns:

Reimplemented from NOX::Abstract::Group.

References dyn_cast().

NOX::Abstract::Group::ReturnType NOX::Thyra::Group::applyJacobianInverse ( Teuchos::ParameterList params,
const NOX::Abstract::Vector input,
NOX::Abstract::Vector result 
) const [virtual]

Applies the inverse of the Jacobian matrix to the given input vector and puts the answer in result.

Computes

\[ v = J^{-1} u, \]

where $J$ is the Jacobian, $u$ is the input vector, and $v$ is the result vector.

The "Tolerance" parameter specifies that the solution should be such that

\[ \frac{\| J v - u \|_2}{\max \{ 1, \|u\|_2\} } < \mbox{Tolerance} \]

Returns:

The parameter "Tolerance" may be added/modified in the list of parameters - this is the ideal solution tolerance for an iterative linear solve.

Reimplemented from NOX::Abstract::Group.

References dyn_cast().

NOX::Abstract::Group::ReturnType NOX::Thyra::Group::applyJacobianInverseMultiVector ( Teuchos::ParameterList params,
const NOX::Abstract::MultiVector input,
NOX::Abstract::MultiVector result 
) const [virtual]

applyJacobianInverse for multiple right-hand sides

The default implementation here calls applyJacobianInverse() for each right hand side serially but should be overloaded if a block solver is available.

Reimplemented from NOX::Abstract::Group.

References Teuchos::dyn_cast(), and NOX::Thyra::MultiVector::getThyraMultiVector().

NOX::Abstract::Group::ReturnType NOX::Thyra::Group::applyJacobianMultiVector ( const NOX::Abstract::MultiVector input,
NOX::Abstract::MultiVector result 
) const [virtual]

applyJacobian for multiple right-hand sides

The default implementation here calls applyJacobian() for each right hand side serially but should be overloaded if a block method is available.

Reimplemented from NOX::Abstract::Group.

References Teuchos::dyn_cast(), NOX::Thyra::MultiVector::getThyraMultiVector(), nonnull(), NOX::Abstract::Group::Ok, Teuchos::RCP< T >::ptr(), and TEUCHOS_TEST_FOR_EXCEPTION.

NOX::Abstract::Group::ReturnType NOX::Thyra::Group::applyJacobianTranspose ( const NOX::Abstract::Vector input,
NOX::Abstract::Vector result 
) const [virtual]

Applies Jacobian-Transpose to the given input vector and puts the answer in the result.

Computes

\[ v = J^T u, \]

where $J$ is the Jacobian, $u$ is the input vector, and $v$ is the result vector.

Returns:

Reimplemented from NOX::Abstract::Group.

References dyn_cast().

NOX::Abstract::Group::ReturnType NOX::Thyra::Group::applyJacobianTransposeMultiVector ( const NOX::Abstract::MultiVector input,
NOX::Abstract::MultiVector result 
) const [virtual]

applyJacobianTranspose for multiple right-hand sides

The default implementation here calls applyJacobianTranspose() for each right hand side serially but should be overloaded if a block method is available.

Reimplemented from NOX::Abstract::Group.

References Teuchos::dyn_cast(), NOX::Abstract::Group::Failed, NOX::Thyra::MultiVector::getThyraMultiVector(), nonnull(), NOX::Abstract::Group::Ok, Teuchos::RCP< T >::ptr(), and TEUCHOS_TEST_FOR_EXCEPTION.

NOX::Abstract::Group::ReturnType NOX::Thyra::Group::applyRightPreconditioning ( bool  useTranspose,
Teuchos::ParameterList params,
const NOX::Abstract::Vector input,
NOX::Abstract::Vector result 
) const [virtual]

Apply right preconditiong to the given input vector.

Let $M$ be a right preconditioner for the Jacobian $J$; in other words, $M$ is a matrix such that

\[ JM \approx I. \]

Compute

\[ u = M^{-1} v, \]

where $u$ is the input vector and $v$ is the result vector.

If useTranspose is true, then the transpose of the preconditioner is applied:

\[ u = {M^{-1}}^T v, \]

The transpose preconditioner is currently only required for Tensor methods.

The "Tolerance" parameter specifies that the solution should be such that

\[ \frac{\| M v - u \|_2}{\max \{ 1, \|u\|_2\} } < \mbox{Tolerance} \]

Returns:

The parameters are from the "Linear %Solver" sublist of the "Direction" sublist that is passed to solver during construction.

Reimplemented from NOX::Abstract::Group.

References NOX::Thyra::Vector::getThyraRCPVector(), is_null(), nonnull(), NOX::Abstract::Group::Ok, and Teuchos::RCP< T >::ptr().

Teuchos::RCP< NOX::Abstract::Group > NOX::Thyra::Group::clone ( NOX::CopyType  type = NOX::DeepCopy) const [virtual]

Create a new Group of the same derived type as this one by cloning this one, and return a ref count pointer to the new group.

If type is NOX::DeepCopy, then we need to create an exact replica of "this". Otherwise, if type is NOX::ShapeCopy, we need only replicate the shape of "this" (only the memory is allocated, the values are not copied into the vectors and Jacobian). Returns NULL if clone is not supported.

Note:
Any shared data should have its ownership transfered to this group from the source for a NOX::DeepCopy.

Implements NOX::Abstract::Group.

Reimplemented in LOCA::Thyra::Group, and LOCA::Thyra::GroupWrapper.

References Teuchos::rcp().

NOX::Abstract::Group::ReturnType NOX::Thyra::Group::computeF ( ) [virtual]

Compute and store F(x).

Note:
It's generally useful to also compute and store the 2-norm of F(x) at this point for later access by the getNormF() function.
Returns:

Implements NOX::Abstract::Group.

Reimplemented in LOCA::Thyra::Group.

References NOX::Abstract::Group::Failed, and NOX::Abstract::Group::Ok.

NOX::Abstract::Group::ReturnType NOX::Thyra::Group::computeGradient ( ) [virtual]

Compute and store gradient.

We can pose the nonlinear equation problem $F(x) = 0$ as an optimization problem as follows:

\[ \min f(x) \equiv \frac{1}{2} \|F(x)\|_2^2. \]

In that case, the gradient (of $f$) is defined as

\[ g \equiv J^T F. \]

Returns:

Reimplemented from NOX::Abstract::Group.

References NOX::Abstract::Group::Failed, nonnull(), NOX::Abstract::Group::Ok, and TEUCHOS_TEST_FOR_EXCEPTION.

NOX::Abstract::Group::ReturnType NOX::Thyra::Group::computeJacobian ( ) [virtual]

Compute and store Jacobian.

Recall that

\[ F(x) = \left[ \begin{array}{c} F_1(x) \\ F_2(x) \\ \vdots \\ F_n(x) \\ \end{array} \right]. \]

The Jacobian is denoted by $J$ and defined by

\[ J_{ij} = \frac{\partial F_i}{\partial x_j} (x). \]

Note:
If this is a shared object, this group should taken ownership of the Jacobian before it computes it.
Returns:

Reimplemented from NOX::Abstract::Group.

Reimplemented in LOCA::Thyra::Group.

References NOX::Abstract::Group::Failed, nonnull(), and NOX::Abstract::Group::Ok.

NOX::Abstract::Group::ReturnType NOX::Thyra::Group::computeNewton ( Teuchos::ParameterList params) [virtual]

Compute the Newton direction, using parameters for the linear solve.

The Newton direction is the solution, s, of

\[ J s = -F. \]

The parameters are from the "Linear %Solver" sublist of the "Direction" sublist that is passed to solver during construction.

The "Tolerance" parameter may be added/modified in the sublist of "Linear Solver" parameters that is passed into this function. The solution should be such that

\[ \frac{\| J s - (-F) \|_2}{\max \{ 1, \|F\|_2\} } < \mbox{Tolerance} \]

Returns:

Reimplemented from NOX::Abstract::Group.

void NOX::Thyra::Group::computeX ( const NOX::Abstract::Group grp,
const NOX::Abstract::Vector d,
double  step 
) [virtual]

Compute x = grp.x + step * d.

Let $x$ denote this group's solution vector. Let $\hat x$ denote the result of grp.getX(). Then set

\[ x = \hat x + \mbox{step} \; d. \]

Note:
This should invalidate the function value, Jacobian, gradient, and Newton direction.
Throw an error if the copy fails.
Returns:
Reference to this object

Implements NOX::Abstract::Group.

Thyra::ModelEvaluatorBase::InArgs< double > & NOX::Thyra::Group::getNonconstInArgs ( )

FOR POWER USERS ONLY! Grab the inargs used by nox and allow the user to change it. Used by pseudo-transient solver to add an x_dot and alpha/beta to the model evaluator call.

double NOX::Thyra::Group::getNormF ( ) const [virtual]

Return 2-norm of F(x).

In other words,

\[ \sqrt{\sum_{i=1}^n F_i^2} \]

Implements NOX::Abstract::Group.

bool NOX::Thyra::Group::isGradient ( ) const [virtual]

Return true if the gradient is valid.

Note:
Default implementation in NOX::Abstract::Group returns false.

Reimplemented from NOX::Abstract::Group.

bool NOX::Thyra::Group::isJacobian ( ) const [virtual]

Return true if the Jacobian is valid.

Note:
Default implementation in NOX::Abstract::Group returns false.

Reimplemented from NOX::Abstract::Group.

References nonnull().

Referenced by Group().

bool NOX::Thyra::Group::isNewton ( ) const [virtual]

Return true if the Newton direction is valid.

Note:
Default implementation in NOX::Abstract::Group returns false.

Reimplemented from NOX::Abstract::Group.

NOX::Abstract::Group & NOX::Thyra::Group::operator= ( const NOX::Abstract::Group source) [virtual]

Copies the source group into this group.

Note:
Any shared data owned by the source should have its ownership transfered to this group. This may result in a secret modification to the source object.

Implements NOX::Abstract::Group.

Reimplemented in LOCA::Thyra::Group, and LOCA::Thyra::GroupWrapper.

Referenced by LOCA::Thyra::Group::operator=().

void NOX::Thyra::Group::setX ( const NOX::Abstract::Vector y) [virtual]

Set the solution vector x to y.

Note:
This should invalidate the function value, Jacobian, gradient, and Newton direction.
Throw an error if the copy fails.
Returns:
Reference to this object

Implements NOX::Abstract::Group.


Member Data Documentation

Teuchos::RCP<const ::Thyra::VectorBase<double> > NOX::Thyra::Group::weight_vec_ [protected]

Optional wieghting vector for function scaling.

This is pulled out of the initial guess vector

Referenced by Group().


The documentation for this class was generated from the following files:
 All Classes Namespaces Files Functions Variables Typedefs Enumerations Enumerator Friends