eigen.texi

来自「math library from gnu」· TEXI 代码 · 共 829 行 · 第 1/3 页

@cindex eigenvalues and eigenvectors
This chapter describes functions for computing eigenvalues and
eigenvectors of matrices.  There are routines for real symmetric,
real nonsymmetric, complex hermitian, real generalized symmetric-definite,
complex generalized hermitian-definite, and real generalized nonsymmetric
eigensystems. Eigenvalues can be computed with or without eigenvectors.
The hermitian and real symmetric matrix algorithms are symmetric bidiagonalization
followed by QR reduction. The nonsymmetric algorithm is the Francis QR
double-shift.  The generalized nonsymmetric algorithm is the QZ method due
to Moler and Stewart.

The functions described in this chapter are declared in the header file
@file{gsl_eigen.h}.

@menu
* Real Symmetric Matrices::     
* Complex Hermitian Matrices::  
* Real Nonsymmetric Matrices::
* Real Generalized Symmetric-Definite Eigensystems::
* Complex Generalized Hermitian-Definite Eigensystems::
* Real Generalized Nonsymmetric Eigensystems::
* Sorting Eigenvalues and Eigenvectors::  
* Eigenvalue and Eigenvector Examples::  
* Eigenvalue and Eigenvector References::  
@end menu

@node Real Symmetric Matrices
@section Real Symmetric Matrices
@cindex symmetric matrix, real, eigensystem
@cindex real symmetric matrix, eigensystem

For real symmetric matrices, the library uses the symmetric
bidiagonalization and QR reduction method.  This is described in Golub
& van Loan, section 8.3.  The computed eigenvalues are accurate to an
absolute accuracy of @math{\epsilon ||A||_2}, where @math{\epsilon} is
the machine precision.

@deftypefun {gsl_eigen_symm_workspace *} gsl_eigen_symm_alloc (const size_t @var{n})
This function allocates a workspace for computing eigenvalues of
@var{n}-by-@var{n} real symmetric matrices.  The size of the workspace
is @math{O(2n)}.
@end deftypefun

@deftypefun void gsl_eigen_symm_free (gsl_eigen_symm_workspace * @var{w})
This function frees the memory associated with the workspace @var{w}.
@end deftypefun

@deftypefun int gsl_eigen_symm (gsl_matrix * @var{A}, gsl_vector * @var{eval}, gsl_eigen_symm_workspace * @var{w})
This function computes the eigenvalues of the real symmetric matrix
@var{A}.  Additional workspace of the appropriate size must be provided
in @var{w}.  The diagonal and lower triangular part of @var{A} are
destroyed during the computation, but the strict upper triangular part
is not referenced.  The eigenvalues are stored in the vector @var{eval}
and are unordered.
@end deftypefun

@deftypefun {gsl_eigen_symmv_workspace *} gsl_eigen_symmv_alloc (const size_t @var{n})
This function allocates a workspace for computing eigenvalues and
eigenvectors of @var{n}-by-@var{n} real symmetric matrices.  The size of
the workspace is @math{O(4n)}.
@end deftypefun

@deftypefun void gsl_eigen_symmv_free (gsl_eigen_symmv_workspace * @var{w})
This function frees the memory associated with the workspace @var{w}.
@end deftypefun

@deftypefun int gsl_eigen_symmv (gsl_matrix * @var{A}, gsl_vector * @var{eval}, gsl_matrix * @var{evec}, gsl_eigen_symmv_workspace * @var{w})
This function computes the eigenvalues and eigenvectors of the real
symmetric matrix @var{A}.  Additional workspace of the appropriate size
must be provided in @var{w}.  The diagonal and lower triangular part of
@var{A} are destroyed during the computation, but the strict upper
triangular part is not referenced.  The eigenvalues are stored in the
vector @var{eval} and are unordered.  The corresponding eigenvectors are
stored in the columns of the matrix @var{evec}.  For example, the
eigenvector in the first column corresponds to the first eigenvalue.
The eigenvectors are guaranteed to be mutually orthogonal and normalised
to unit magnitude.
@end deftypefun

@node Complex Hermitian Matrices
@section Complex Hermitian Matrices

@cindex hermitian matrix, complex, eigensystem
@cindex complex hermitian matrix, eigensystem

@deftypefun {gsl_eigen_herm_workspace *} gsl_eigen_herm_alloc (const size_t @var{n})
This function allocates a workspace for computing eigenvalues of
@var{n}-by-@var{n} complex hermitian matrices.  The size of the workspace
is @math{O(3n)}.
@end deftypefun

@deftypefun void gsl_eigen_herm_free (gsl_eigen_herm_workspace * @var{w})
This function frees the memory associated with the workspace @var{w}.
@end deftypefun

@deftypefun int gsl_eigen_herm (gsl_matrix_complex * @var{A}, gsl_vector * @var{eval}, gsl_eigen_herm_workspace * @var{w})
This function computes the eigenvalues of the complex hermitian matrix
@var{A}.  Additional workspace of the appropriate size must be provided
in @var{w}.  The diagonal and lower triangular part of @var{A} are
destroyed during the computation, but the strict upper triangular part
is not referenced.  The imaginary parts of the diagonal are assumed to be
zero and are not referenced. The eigenvalues are stored in the vector
@var{eval} and are unordered.
@end deftypefun

@deftypefun {gsl_eigen_hermv_workspace *} gsl_eigen_hermv_alloc (const size_t @var{n})
This function allocates a workspace for computing eigenvalues and
eigenvectors of @var{n}-by-@var{n} complex hermitian matrices.  The size of
the workspace is @math{O(5n)}.
@end deftypefun

@deftypefun void gsl_eigen_hermv_free (gsl_eigen_hermv_workspace * @var{w})
This function frees the memory associated with the workspace @var{w}.
@end deftypefun

@deftypefun int gsl_eigen_hermv (gsl_matrix_complex * @var{A}, gsl_vector * @var{eval}, gsl_matrix_complex * @var{evec}, gsl_eigen_hermv_workspace * @var{w})
This function computes the eigenvalues and eigenvectors of the complex
hermitian matrix @var{A}.  Additional workspace of the appropriate size
must be provided in @var{w}.  The diagonal and lower triangular part of
@var{A} are destroyed during the computation, but the strict upper
triangular part is not referenced. The imaginary parts of the diagonal
are assumed to be zero and are not referenced.  The eigenvalues are
stored in the vector @var{eval} and are unordered.  The corresponding
complex eigenvectors are stored in the columns of the matrix @var{evec}.
For example, the eigenvector in the first column corresponds to the
first eigenvalue.  The eigenvectors are guaranteed to be mutually
orthogonal and normalised to unit magnitude.
@end deftypefun

@node Real Nonsymmetric Matrices
@section Real Nonsymmetric Matrices
@cindex nonsymmetric matrix, real, eigensystem
@cindex real nonsymmetric matrix, eigensystem

The solution of the real nonsymmetric eigensystem problem for a
matrix @math{A} involves computing the Schur decomposition
@tex
\beforedisplay
$$
A = Z T Z^T
$$
\afterdisplay
@end tex
@ifinfo

@example
A = Z T Z^T
@end example

@end ifinfo
where @math{Z} is an orthogonal matrix of Schur vectors and @math{T},
the Schur form, is quasi upper triangular with diagonal
@math{1}-by-@math{1} blocks which are real eigenvalues of @math{A}, and
diagonal @math{2}-by-@math{2} blocks whose eigenvalues are complex
conjugate eigenvalues of @math{A}. The algorithm used is the double
shift Francis method.

@deftypefun {gsl_eigen_nonsymm_workspace *} gsl_eigen_nonsymm_alloc (const size_t @var{n})
This function allocates a workspace for computing eigenvalues of
@var{n}-by-@var{n} real nonsymmetric matrices. The size of the workspace
is @math{O(2n)}.
@end deftypefun

@deftypefun void gsl_eigen_nonsymm_free (gsl_eigen_nonsymm_workspace * @var{w})
This function frees the memory associated with the workspace @var{w}.
@end deftypefun

@deftypefun void gsl_eigen_nonsymm_params (const int @var{compute_t}, const int @var{balance}, gsl_eigen_nonsymm_workspace * @var{w})
This function sets some parameters which determine how the eigenvalue
problem is solved in subsequent calls to @code{gsl_eigen_nonsymm}.

If @var{compute_t} is set to 1, the full Schur form @math{T} will be
computed by @code{gsl_eigen_nonsymm}. If it is set to 0,
@math{T} will not be computed (this is the default setting). Computing
the full Schur form @math{T} requires approximately 1.5-2 times the
number of flops.

If @var{balance} is set to 1, a balancing transformation is applied
to the matrix prior to computing eigenvalues. This transformation is
designed to make the rows and columns of the matrix have comparable
norms, and can result in more accurate eigenvalues for matrices
whose entries vary widely in magnitude. See @ref{Balancing} for more
information. Note that the balancing transformation does not preserve
the orthogonality of the Schur vectors, so if you wish to compute the
Schur vectors with @code{gsl_eigen_nonsymm_Z} you will obtain the Schur
vectors of the balanced matrix instead of the original matrix. The
relationship will be
@tex
\beforedisplay
$$
T = Q^t D^{-1} A D Q
$$
\afterdisplay
@end tex
@ifinfo

@example
T = Q^t D^(-1) A D Q
@end example

@end ifinfo
@noindent
where @var{Q} is the matrix of Schur vectors for the balanced matrix, and
@var{D} is the balancing transformation. Then @code{gsl_eigen_nonsymm_Z}
will compute a matrix @var{Z} which satisfies
@tex
\beforedisplay
$$
T = Z^{-1} A Z
$$
\afterdisplay
@end tex
@ifinfo

@example
T = Z^(-1) A Z
@end example

@end ifinfo
@noindent
with @math{Z = D Q}. Note that @var{Z} will not be orthogonal. For
this reason, balancing is not performed by default.
@end deftypefun

@deftypefun int gsl_eigen_nonsymm (gsl_matrix * @var{A}, gsl_vector_complex * @var{eval}, gsl_eigen_nonsymm_workspace * @var{w})
This function computes the eigenvalues of the real nonsymmetric matrix
@var{A} and stores them in the vector @var{eval}. If @math{T} is
desired, it is stored in the upper portion of @var{A} on output.
Otherwise, on output, the diagonal of @var{A} will contain the
@math{1}-by-@math{1} real eigenvalues and @math{2}-by-@math{2}
complex conjugate eigenvalue systems, and the rest of @var{A} is
destroyed. In rare cases, this function may fail to find all
eigenvalues. If this happens, an error code is returned
and the number of converged eigenvalues is stored in @code{w->n_evals}.
The converged eigenvalues are stored in the beginning of @var{eval}.
@end deftypefun

@deftypefun int gsl_eigen_nonsymm_Z (gsl_matrix * @var{A}, gsl_vector_complex * @var{eval}, gsl_matrix * @var{Z}, gsl_eigen_nonsymm_workspace * @var{w})
This function is identical to @code{gsl_eigen_nonsymm} except it also
computes the Schur vectors and stores them into @var{Z}.
@end deftypefun

@deftypefun {gsl_eigen_nonsymmv_workspace *} gsl_eigen_nonsymmv_alloc (const size_t @var{n})
This function allocates a workspace for computing eigenvalues and
eigenvectors of @var{n}-by-@var{n} real nonsymmetric matrices. The
size of the workspace is @math{O(5n)}.
@end deftypefun

@deftypefun void gsl_eigen_nonsymmv_free (gsl_eigen_nonsymmv_workspace * @var{w})
This function frees the memory associated with the workspace @var{w}.
@end deftypefun

@deftypefun int gsl_eigen_nonsymmv (gsl_matrix * @var{A}, gsl_vector_complex * @var{eval}, gsl_matrix_complex * @var{evec}, gsl_eigen_nonsymmv_workspace * @var{w})
This function computes eigenvalues and right eigenvectors of the
@var{n}-by-@var{n} real nonsymmetric matrix @var{A}. It first calls
@code{gsl_eigen_nonsymm} to compute the eigenvalues, Schur form @math{T}, and
Schur vectors. Then it finds eigenvectors of @math{T} and backtransforms
them using the Schur vectors. The Schur vectors are destroyed in the
process, but can be saved by using @code{gsl_eigen_nonsymmv_Z}. The
computed eigenvectors are normalized to have unit magnitude. On
output, the upper portion of @var{A} contains the Schur form
@math{T}. If @code{gsl_eigen_nonsymm} fails, no eigenvectors are
computed, and an error code is returned.
@end deftypefun

@deftypefun int gsl_eigen_nonsymmv_Z (gsl_matrix * @var{A}, gsl_vector_complex * @var{eval}, gsl_matrix_complex * @var{evec}, gsl_matrix * @var{Z}, gsl_eigen_nonsymmv_workspace * @var{w})
This function is identical to @code{gsl_eigen_nonsymmv} except it also saves
the Schur vectors into @var{Z}.
@end deftypefun

@node Real Generalized Symmetric-Definite Eigensystems
@section Real Generalized Symmetric-Definite Eigensystems
@cindex generalized symmetric eigensystems

The real generalized symmetric-definite eigenvalue problem is to find
eigenvalues @math{\lambda} and eigenvectors @math{x} such that
@tex