Kaskade::Bridge::Functional< FunctionalImpl, DomainImpl, ImageImpl > Class Template Reference

Object that represents the linearization of a nonlinear functional, implements AbstractLineariztion. More...

#include <newton_bridge.hh>

Detailed Description

template<typename FunctionalImpl, typename DomainImpl = typename FunctionalImpl::AnsatzVars::VariableSet, typename ImageImpl = DomainImpl>
class Kaskade::Bridge::Functional< FunctionalImpl, DomainImpl, ImageImpl >

Object that represents the linearization of a nonlinear functional, implements AbstractLineariztion.

Can be connected to an AdaptiveGrid to be able to know, when the grid changes Emits a signal if it is invalidated. This can be used by a solver to flush data

Currently there are two implementations available with an automatic switch depending on the type of vector used. This switch is implemented via LinearizationTraits:

The first implementation is used in case of a Dune::BlockVector<T>. Then the creation of a linear system is forwarded to the functional

The second implementation is used in case of a VariableSet. Here a VariationalFunctionalAssembler is used to create the linear system.

Implementations have to provide the following routines Example: DuneLinearization

• RT getValue()
• void getRHSBlocks(std::vector<RT>& rhs, int rbegin, int rend) const
• void getMatrixBlocks(MatrixAsTriplet<RT>& mat, int rbegin, int rend, int colbegin, int colend) const
• AbstractFunctionSpaceElement const& getOrigin()
• int size()
• void flush()
• int nColBlocks();
• int nRowBlocks();
• int cols(int cbegin, int cend) const
• int rows(int rbegin, int rend) const

Linearizations can be defined to set the right hand side to 0 everywhere, except in [rhsstart,rhsfinal) this is useful on the optimization context Linear Solver. Can get a signal from Linearization in order to flush its data, implements AbstractLinearSolver Implementation must support the following member functions:

• getRelativeAccuracy() : relative Accuracy of the Solution
• solve(std::vector<std::vector<double> >& c, SparseLinearSystem& lin) : solve a linear system
• resolve(std::vector<std::vector<double> >& c, SparseLinearSystem& lin) : solve a linear system, using an old factorization
• onChangedLinearization() : do something, if the Linearization has changed, for example flush memory

There is one pure virtual function to be overloaded Bridge class for Functionals. Its most prominent task is to create linearizations, implements AbstractFunctional

Definition at line 652 of file newton_bridge.hh.

Inheritance diagram for Kaskade::Bridge::Functional< FunctionalImpl, DomainImpl, ImageImpl >:

Public Types
typedef FunctionalImpl::Scalar	Scalar

Public Member Functions
	Functional (FunctionalImpl *impl)
	Functional (std::unique_ptr< FunctionalImpl > &&imp)
template<typename... ConstructorArguments>
	Functional (const ConstructorArguments &... args)
std::unique_ptr< AbstractLinearization >	getLinearization (AbstractFunctionSpaceElement const &x) const
virtual std::unique_ptr< AbstractFunctionSpaceElement >	getImageVector (AbstractFunctionSpaceElement const &x) const
virtual bool	inDomain (AbstractFunctionSpaceElement const &x) const
virtual double	evaluate (AbstractFunctionSpaceElement const &x) const

Protected Attributes
std::unique_ptr< FunctionalImpl >	myImplementation

Member Typedef Documentation

Scalar

template<typename FunctionalImpl , typename DomainImpl = typename FunctionalImpl::AnsatzVars::VariableSet, typename ImageImpl = DomainImpl>

typedef FunctionalImpl::Scalar Kaskade::Bridge::Functional< FunctionalImpl, DomainImpl, ImageImpl >::Scalar

Definition at line 655 of file newton_bridge.hh.

Constructor & Destructor Documentation

Functional() [1/3]

template<typename FunctionalImpl , typename DomainImpl = typename FunctionalImpl::AnsatzVars::VariableSet, typename ImageImpl = DomainImpl>

Kaskade::Bridge::Functional< FunctionalImpl, DomainImpl, ImageImpl >::Functional ( FunctionalImpl *  impl )

inlineexplicit

Definition at line 658 of file newton_bridge.hh.

Functional() [2/3]

template<typename FunctionalImpl , typename DomainImpl = typename FunctionalImpl::AnsatzVars::VariableSet, typename ImageImpl = DomainImpl>

Kaskade::Bridge::Functional< FunctionalImpl, DomainImpl, ImageImpl >::Functional ( std::unique_ptr< FunctionalImpl > &&  imp )

inlineexplicit

Definition at line 660 of file newton_bridge.hh.

Functional() [3/3]

template<typename FunctionalImpl , typename DomainImpl = typename FunctionalImpl::AnsatzVars::VariableSet, typename ImageImpl = DomainImpl>

template<typename... ConstructorArguments>

Kaskade::Bridge::Functional< FunctionalImpl, DomainImpl, ImageImpl >::Functional ( const ConstructorArguments &...  args )

inlineexplicit

Definition at line 663 of file newton_bridge.hh.

Member Function Documentation

evaluate()

virtual double Kaskade::AbstractFunctional::evaluate ( AbstractFunctionSpaceElement const &  x ) const

inlinevirtualinherited

Definition at line 226 of file abstract_interface.hh.

getImageVector()

template<typename FunctionalImpl , typename DomainImpl = typename FunctionalImpl::AnsatzVars::VariableSet, typename ImageImpl = DomainImpl>

virtual std::unique_ptr< AbstractFunctionSpaceElement > Kaskade::Bridge::Functional< FunctionalImpl, DomainImpl, ImageImpl >::getImageVector ( AbstractFunctionSpaceElement const &  x ) const

inlinevirtual

Implements Kaskade::AbstractFunctional.

Definition at line 671 of file newton_bridge.hh.

getLinearization()

template<typename FunctionalImpl , typename DomainImpl = typename FunctionalImpl::AnsatzVars::VariableSet, typename ImageImpl = DomainImpl>

std::unique_ptr< AbstractLinearization > Kaskade::Bridge::Functional< FunctionalImpl, DomainImpl, ImageImpl >::getLinearization ( AbstractFunctionSpaceElement const &  x ) const

inlinevirtual

Implements Kaskade::AbstractFunctional.

Definition at line 665 of file newton_bridge.hh.

inDomain()

template<typename FunctionalImpl , typename DomainImpl = typename FunctionalImpl::AnsatzVars::VariableSet, typename ImageImpl = DomainImpl>

virtual bool Kaskade::Bridge::Functional< FunctionalImpl, DomainImpl, ImageImpl >::inDomain ( AbstractFunctionSpaceElement const &  x ) const

inlinevirtual

Reimplemented from Kaskade::AbstractFunctional.

Definition at line 677 of file newton_bridge.hh.

Member Data Documentation

myImplementation

template<typename FunctionalImpl , typename DomainImpl = typename FunctionalImpl::AnsatzVars::VariableSet, typename ImageImpl = DomainImpl>

std::unique_ptr<FunctionalImpl> Kaskade::Bridge::Functional< FunctionalImpl, DomainImpl, ImageImpl >::myImplementation

protected

Definition at line 682 of file newton_bridge.hh.

Referenced by Kaskade::Bridge::Functional< FunctionalImpl, DomainImpl, ImageImpl >::getLinearization(), and Kaskade::Bridge::Functional< FunctionalImpl, DomainImpl, ImageImpl >::inDomain().

---

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

• newton_bridge.hh