Implements a conjugate gradient method as sparse linear algebra solver. More...

#include <conjugate_gradient_method.hpp>

Inheritance diagram for ConjugateGradientMethod:

Public Member Functions
virtual SparseSolver *	clone () const =0

	ConjugateGradientMethod ()

	ConjugateGradientMethod (const ConjugateGradientMethod &arg)

virtual returnValue	getX (double *x_)

virtual returnValue	setDimension (const int &n)

virtual returnValue	setIndices (const int rowIdx_, const int colIdx_)=0

virtual returnValue	setMatrix (double *A_)

virtual returnValue	setNumberOfEntries (const int &nDense_)

virtual returnValue	setPrintLevel (PrintLevel printLevel_)

virtual returnValue	setTolerance (double TOL_)

virtual returnValue	solve (double *b)

virtual	~ConjugateGradientMethod ()

Public Member Functions inherited from SparseSolver
virtual returnValue	factorize ()=0

virtual int_t	getNegativeEigenvalues ()

virtual int_t	getRank ()

virtual returnValue	getZeroPivots (int_t *&zeroPivots)

virtual SparseSolver &	operator= (const SparseSolver &rhs)

virtual returnValue	reset ()

virtual returnValue	setMatrixData (int_t dim, int_t numNonzeros, const int_t const airn, const int_t const acjn, const real_t *const avals)=0

virtual returnValue	solve (int_t dim, const real_t const rhs, real_t const sol)=0

virtual returnValue	solveTranspose (double *b)

	SparseSolver ()

	SparseSolver (const SparseSolver &rhs)

	SparseSolver ()

virtual	~SparseSolver ()

virtual	~SparseSolver ()

Protected Member Functions
virtual returnValue	applyInversePreconditioner (double *x_)=0

virtual returnValue	applyPreconditioner (double *b)=0

virtual returnValue	computePreconditioner (double *A_)=0

virtual void	multiply (double xx, double result)=0

double	scalarProduct (double aa, double bb)

Protected Member Functions inherited from SparseSolver
returnValue	clear ()

returnValue	copy (const SparseSolver &rhs)

Protected Attributes
double *	A

double *	condScale

int	dim

int	nDense

double *	norm2

double **	p

int	pCounter

PrintLevel	printLevel

double *	r

double	TOL

double *	x

Detailed Description

Implements a conjugate gradient method as sparse linear algebra solver.

The class ConjugateGradientMethod implements a special sparse
linear algebra solver. After the application of an preconditioner
an iterative conjugate basis of the sparse data matrix A is
computed. The algotithm stops if the required tolerance is achieved.

Author: Boris Houska, Hans Joachim Ferreau

Definition at line 57 of file conjugate_gradient_method.hpp.

Constructor & Destructor Documentation

BEGIN_NAMESPACE_ACADO ConjugateGradientMethod::ConjugateGradientMethod ( )

Default constructor.

Definition at line 48 of file conjugate_gradient_method.cpp.

ConjugateGradientMethod::ConjugateGradientMethod ( const ConjugateGradientMethod & arg )

Copy constructor (deep copy).

Definition at line 64 of file conjugate_gradient_method.cpp.

ConjugateGradientMethod::~ConjugateGradientMethod ( )

virtual

Destructor.

Definition at line 126 of file conjugate_gradient_method.cpp.

Member Function Documentation

virtual returnValue ConjugateGradientMethod::applyInversePreconditioner ( double * x_ )

protectedpure virtual

Applies the inverse of the preconditioner to the vector x (only internal use)

Implemented in SymmetricConjugateGradientMethod, and NormalConjugateGradientMethod.

virtual returnValue ConjugateGradientMethod::applyPreconditioner ( double * b )

protectedpure virtual

Applies the preconditioner to the vector b (only internal use)

Implemented in SymmetricConjugateGradientMethod, and NormalConjugateGradientMethod.

virtual SparseSolver* ConjugateGradientMethod::clone ( ) const

pure virtual

Clone operator (deep copy).

Implements SparseSolver.

Implemented in SymmetricConjugateGradientMethod, and NormalConjugateGradientMethod.

virtual returnValue ConjugateGradientMethod::computePreconditioner ( double * A_ )

protectedpure virtual

Computes the preconditioner and Applies it to the input matrix.

Implemented in SymmetricConjugateGradientMethod, and NormalConjugateGradientMethod.

returnValue ConjugateGradientMethod::getX ( double * x_ )

virtual

Returns the solution of the equation A*x = b if solved.

Returns: SUCCESSFUL_RETURN

Implements SparseSolver.

Definition at line 329 of file conjugate_gradient_method.cpp.

virtual void ConjugateGradientMethod::multiply	(	double *	xx,
		double *	result
	)

protectedpure virtual

Evaluates the matrix-vector product result = A*xx efficiently. (only internal use)

Implemented in SymmetricConjugateGradientMethod, and NormalConjugateGradientMethod.

double ConjugateGradientMethod::scalarProduct	(	double *	aa,
		double *	bb
	)

protected

Returns the scalar product of aa and bb. (only internal use)

Definition at line 355 of file conjugate_gradient_method.cpp.

returnValue ConjugateGradientMethod::setDimension ( const int & n )

virtual

Defines the dimension n of A R^{n n}

Returns: SUCCESSFUL_RETURN

Implements SparseSolver.

Definition at line 252 of file conjugate_gradient_method.cpp.

virtual returnValue ConjugateGradientMethod::setIndices	(	const int *	rowIdx_,
		const int *	colIdx_
	)

pure virtual

Sets an index list containing the positions of the
non-zero elements in the matrix A.

Implements SparseSolver.

Implemented in SymmetricConjugateGradientMethod, and NormalConjugateGradientMethod.

returnValue ConjugateGradientMethod::setMatrix ( double * A_ )

virtual

Sets the non-zero elements of the matrix A. The double* A
is assumed to contain nDense entries corresponding to
non-zero elements of A.

Implements SparseSolver.

Definition at line 315 of file conjugate_gradient_method.cpp.

returnValue ConjugateGradientMethod::setNumberOfEntries ( const int & nDense_ )

virtual

Defines the number of non-zero elements in the
matrix A

Returns: SUCCESSFUL_RETURN

Implements SparseSolver.

Definition at line 300 of file conjugate_gradient_method.cpp.

returnValue ConjugateGradientMethod::setPrintLevel ( PrintLevel printLevel_ )

virtual

Sets the print level.

Returns: SUCCESSFUL_RETURN

Implements SparseSolver.

Definition at line 342 of file conjugate_gradient_method.cpp.

returnValue ConjugateGradientMethod::setTolerance ( double TOL_ )

virtual

Sets the required tolerance (accuracy) for the solution of
the linear equation. For large tolerances an iterative
algorithm might converge earlier.

Requires || A*x - b || <= TOL

The norm || . || is possibly scaled by a preconditioner.