SVDSetConvergenceTest#

Specifies how to compute the error estimate used in the convergence test.

Synopsis#

#include "slepcsvd.h" 
PetscErrorCode SVDSetConvergenceTest(SVD svd,SVDConv conv)

Logically Collective

Input Parameters#

  • svd - the singular value solver context

  • conv - the type of convergence test

Options Database Keys#

  • -svd_conv_abs - Sets the absolute convergence test

  • -svd_conv_rel - Sets the convergence test relative to the singular value

  • -svd_conv_norm - Sets the convergence test relative to the matrix norm

  • -svd_conv_maxit - Forces the maximum number of iterations as set by -svd_max_it

  • -svd_conv_user - Selects the user-defined convergence test

Notes#

The parameter ‘conv’ can have one of these values

  • SVD_CONV_ABS - absolute error ||r||

  • SVD_CONV_REL - error relative to the singular value sigma, ||r||/sigma

  • SVD_CONV_NORM - error relative to the matrix norms, ||r||/||Z||, with Z=A or Z=[A;B]

  • SVD_CONV_MAXIT - no convergence until maximum number of iterations has been reached

  • SVD_CONV_USER - function set by SVDSetConvergenceTestFunction()

The default in standard SVD is SVD_CONV_REL, while in GSVD the default is SVD_CONV_NORM.

See Also#

SVD: Singular Value Decomposition, SVDGetConvergenceTest(), SVDSetConvergenceTestFunction(), SVDSetStoppingTest(), SVDConv

Level#

intermediate

Location#

src/svd/interface/svdopts.c

Index of all SVD routines Table of Contents for all manual pages Index of all manual pages