hide long namesshow long names
hide short namesshow short names
Integer type:  int32  int64  nag_int  show int32  show int32  show int64  show int64  show nag_int  show nag_int

PDF version (NAG web site, 64-bit version, 64-bit version)
Chapter Contents
Chapter Introduction
NAG Toolbox

NAG Toolbox: nag_eigen_real_symm_sparse_arnoldi (f02fk)

 Contents

    1  Purpose
    2  Syntax
    7  Accuracy
    9  Example

Purpose

nag_eigen_real_symm_sparse_arnoldi (f02fk) computes selected eigenvalues and eigenvectors of a real sparse symmetric matrix.

Syntax

[nconv, w, v, resid, user, ifail] = f02fk(n, a, irow, icol, nev, ncv, sigma, monit, option, 'nz', nz, 'user', user)
[nconv, w, v, resid, user, ifail] = nag_eigen_real_symm_sparse_arnoldi(n, a, irow, icol, nev, ncv, sigma, monit, option, 'nz', nz, 'user', user)

Description

nag_eigen_real_symm_sparse_arnoldi (f02fk) computes selected eigenvalues and the corresponding right eigenvectors of a real sparse symmetric matrix A:
Avi = λi vi .  
A specified number, nev, of eigenvalues λi, or the shifted inverses μi=1/λi-σ, may be selected either by largest or smallest modulus, largest or smallest value, or, largest and smallest values (both ends). Convergence is generally faster when selecting larger eigenvalues, smaller eigenvalues can always be selected by choosing a zero inverse shift (σ=0.0). When eigenvalues closest to a given value are required then the shifted inverses of largest magnitude should be selected with shift equal to the required value.
The sparse matrix A is stored in symmetric coordinate storage (SCS) format. See Symmetric coordinate storage (SCS) format in the F11 Chapter Introduction.
nag_eigen_real_symm_sparse_arnoldi (f02fk) uses an implicitly restarted Arnoldi (Lanczos) iterative method to converge approximations to a set of required eigenvalues and corresponding eigenvectors. Further algorithmic information is given in Further Comments while a fuller discussion is provided in the F12 Chapter Introduction. If shifts are to be performed then operations using shifted inverse matrices are performed using a direct sparse solver.

References

Golub G H and Van Loan C F (1996) Matrix Computations (3rd Edition) Johns Hopkins University Press, Baltimore
Lehoucq R B, Sorensen D C and Yang C (1998) ARPACK Users' Guide: Solution of Large-scale Eigenvalue Problems with Implicitly Restarted Arnoldi Methods SIAM, Philidelphia

Parameters

Compulsory Input Parameters

1:     n int64int32nag_int scalar
n, the order of the matrix A.
Constraint: n>0.
2:     anz – double array
The array of nonzero elements of the lower triangular part of the n by n symmetric matrix A.
3:     irownz int64int32nag_int array
4:     icolnz int64int32nag_int array
The row and column indices of the elements supplied in array a.
If irowk=i and icolk=j then Aij is stored in ak. irow does not need to be ordered, an internal call to nag_sparse_real_symm_sort (f11zb) forces the correct ordering.
Constraint:
irow and icol must satisfy these constraints:
1irowin and 1icoliirowi, for i=1,2,,nz.
5:     nev int64int32nag_int scalar
The number of eigenvalues to be computed.
Constraint: 0<nev<n-1.
6:     ncv int64int32nag_int scalar
The dimension of the array w. the number of Arnoldi basis vectors to use during the computation.
At present there is no a priori analysis to guide the selection of ncv relative to nev. However, it is recommended that ncv2×nev+1. If many problems of the same type are to be solved, you should experiment with increasing ncv while keeping nev fixed for a given test problem. This will usually decrease the required number of matrix-vector operations but it also increases the work and storage required to maintain the orthogonal basis vectors. The optimal ‘cross-over’ with respect to computation time is problem dependent and must be determined empirically.
Constraint: nev<ncvn.
7:     sigma – double scalar
If the Shifted Inverse mode has been selected then sigma contains the real shift used; otherwise sigma is not referenced. This mode can be selected by setting the appropriate options in the user-supplied function option.
8:     monit – function handle or string containing name of m-file
monit is used to monitor the progress of nag_eigen_real_symm_sparse_arnoldi (f02fk). monit may be the dummy function nag_eigen_arnoldi_monit_symm (f02fkz) if no monitoring is actually required. (nag_eigen_arnoldi_monit_symm (f02fkz) is included in the NAG Toolbox.) monit is called after the solution of each eigenvalue sub-problem and also just prior to return from nag_eigen_real_symm_sparse_arnoldi (f02fk).
[istat, user] = monit(ncv, niter, nconv, w, rzest, istat, user)

Input Parameters

1:     ncv int64int32nag_int scalar
The dimension of the arrays w and rzest. The number of Arnoldi basis vectors used during the computation.
2:     niter int64int32nag_int scalar
The number of the current Arnoldi iteration.
3:     nconv int64int32nag_int scalar
The number of converged eigenvalues so far.
4:     wncv – double array
The first nconv elements of w contain the converged approximate eigenvalues.
5:     rzestncv – double array
The first nconv elements of rzest contain the Ritz estimates (error bounds) on the converged approximate eigenvalues.
6:     istat int64int32nag_int scalar
Set to zero.
7:     user – Any MATLAB object
monit is called from nag_eigen_real_symm_sparse_arnoldi (f02fk) with the object supplied to nag_eigen_real_symm_sparse_arnoldi (f02fk).

Output Parameters

1:     istat int64int32nag_int scalar
If set to a nonzero value nag_eigen_real_symm_sparse_arnoldi (f02fk) returns immediately with ifail=9.
2:     user – Any MATLAB object
9:     option – function handle or string containing name of m-file
You can supply non-default options to the Arnoldi eigensolver by repeated calls to nag_sparseig_real_symm_option (f12fd) from within option. (Please note that it is only necessary to call nag_sparseig_real_symm_option (f12fd); no call to nag_sparseig_real_symm_init (f12fa) is required from within option.) For example, you can set the mode to Shifted Inverse, you can increase the Iteration Limit beyond its default and you can print varying levels of detail on the iterative process using Print Level.
If only the default options (including that the eigenvalues of largest magnitude are sought) are to be used then option may be the dummy function nag_eigen_arnoldi_option (f02eky) (nag_eigen_arnoldi_option (f02eky) is included in the NAG Toolbox). See Example for an example of using option to set some non-default options.
[icomm, comm, istat, user] = option(icomm, comm, istat, user)

Input Parameters

1:     icomm: int64int32nag_int array
Contains details of the default option set. This array must be passed as argument icomm in any call to nag_sparseig_real_symm_option (f12fd).
2:     comm: – double array
Contains details of the default option set. This array must be passed as argument comm in any call to nag_sparseig_real_symm_option (f12fd).
3:     istat int64int32nag_int scalar
Set to zero.
4:     user – Any MATLAB object
option is called from nag_eigen_real_symm_sparse_arnoldi (f02fk) with the object supplied to nag_eigen_real_symm_sparse_arnoldi (f02fk).

Output Parameters

1:     icomm: int64int32nag_int array
Contains data on the current options set which may be altered from the default set via calls to nag_sparseig_real_symm_option (f12fd).
2:     comm: – double array
Contains data on the current options set which may be altered from the default set via calls to nag_sparseig_real_symm_option (f12fd).
3:     istat int64int32nag_int scalar
If set to a nonzero value nag_eigen_real_symm_sparse_arnoldi (f02fk) returns immediately with ifail=10.
4:     user – Any MATLAB object

Optional Input Parameters

1:     nz int64int32nag_int scalar
Default: the dimension of the arrays irow, icol and the dimension of the array a. (An error is raised if these dimensions are not equal.)
The dimension of the array a.the number of nonzero elements in the lower triangular part of the matrix A.
Constraint: 1nzn×n+1/2.
2:     user – Any MATLAB object
user is not used by nag_eigen_real_symm_sparse_arnoldi (f02fk), but is passed to monit and option. Note that for large objects it may be more efficient to use a global variable which is accessible from the m-files than to use user.

Output Parameters

1:     nconv int64int32nag_int scalar
The number of converged approximations to the selected eigenvalues. On successful exit, this will normally be nev.
2:     wncv – double array
The first nconv elements contain the converged approximations to the selected eigenvalues.
3:     vldv: – double array
The first dimension of the array v will be n.
The second dimension of the array v will be ncv.
Contains the eigenvectors associated with the eigenvalue λi, for i=1,2,,nconv (stored in w). For eigenvalue, λj, the corresponding eigenvector is stored in vij, for i=1,2,,n.
4:     residnev – double array
The residual A wi - λi wi 2 for the estimates to the eigenpair λi and wi is returned in residi, for i=1,2,,nconv.
5:     user – Any MATLAB object
6:     ifail int64int32nag_int scalar
ifail=0 unless the function detects an error (see Error Indicators and Warnings).

Error Indicators and Warnings

Errors or warnings detected by the function:
   ifail=1
Constraint: n>0.
   ifail=2
Constraint: nz>0.
Constraint: nzn×n+1/2.
   ifail=4
Constraint: 1irowin.
   ifail=5
Constraint: 1icoliirowi.
   ifail=6
Constraint: nev<n-1.
Constraint: nev>0.
   ifail=7
Constraint: ncv>nev.
Constraint: ncvn.
   ifail=8
On entry, the matrix A-σI is numerically singular and could not be inverted. Try perturbing the value of σ.
   ifail=9
User requested termination in monit.
   ifail=10
User requested termination in option.
   ifail=14
Constraint: ldvn.
   ifail=20
The maximum number of iterations, through the optional parameter Iteration Limit, has been set to a non-positive value.
   ifail=21
The option Both Ends has been set but only 1 eigenvalue is requested.
   ifail=22
The maximum number of iterations has been reached.
   ifail=30
A serious error, code _, has occurred in an internal call to _. Check all function calls and array sizes. If the call is correct then please contact NAG for assistance.
   ifail=-99
An unexpected error has been triggered by this routine. Please contact NAG.
   ifail=-399
Your licence key may have expired or may not have been installed correctly.
   ifail=-999
Dynamic memory allocation failed.

Accuracy

The relative accuracy of a Ritz value (eigenvalue approximation), λ, is considered acceptable if its Ritz estimate Tolerance×λ. The default value for Tolerance is the machine precision given by nag_machine_precision (x02aj). The Ritz estimates are available via the monit function at each iteration in the Arnoldi process, or can be printed by setting option Print Level to a positive value.

Further Comments

nag_eigen_real_symm_sparse_arnoldi (f02fk) calls functions based on the ARPACK suite in Chapter F12. These functions use an implicitly restarted Lanczos iterative method to converge to approximations to a set of required eigenvalues (see the F12 Chapter Introduction).
In the default Regular mode, only matrix-vector multiplications are performed using the sparse matrix A during the Lanczos process; nag_sparse_real_symm_matvec (f11xe) can be used to perform this task. Each iteration is therefore cheap computationally, relative to the alternative, Shifted Inverse, mode described below. It is most efficient (i.e., the total number of iterations required is small) when the eigenvalues of largest magnitude are sought and these are distinct.
Although there is an option for returning the smallest eigenvalues using this mode (see Smallest Magnitude option), the number of iterations required for convergence will be far greater or the method may not converge at all. However, where convergence is achieved, Regular mode may still prove to be the most efficient since no inversions are required. Where smallest eigenvalues are sought and Regular mode is not suitable, or eigenvalues close to a given real value are sought, the Shifted Inverse mode should be used.
If the Shifted Inverse mode is used (via a call to nag_sparseig_real_symm_option (f12fd) in option) then the matrix A-σI is used in linear system solves by the Lanczos process. This is first factorized internally using a direct sparse LDLT factorization under the assumption that the matrix is indefinite. If the factorization determines that the matrix is numerically singular then the function exits with an error. In this situation it is normally sufficient to perturb σ by a small amount and call nag_eigen_real_symm_sparse_arnoldi (f02fk) again. After successful factorization, subsequent solves are performed by backsubstitution using the sparse factorization.
Finally, nag_eigen_real_symm_sparse_arnoldi (f02fk) transforms the eigenvectors. Each eigenvector w is normalized so that w2=1.
The monitoring function monit provides some basic information on the convergence of the Lanczos iterations. Much greater levels of detail on the Lanczos process are available via option Print Level. If this is set to a positive value then information will be printed, by default, to standard output. The destination of monitoring information can be changed using the Monitoring option.

Example

This example solves Ax = λx  in Shifted Inverse mode, where A  is obtained from the standard central difference discretization of the one-dimensional Laplacian operator 2u x2  on 0,1 , with zero Dirichlet boundary conditions.
function f02fk_example


fprintf('f02fk example results\n\n');


% This example demonstrates the use of f02fk to evaluate a number of
% eigenvalues of a sparse symmetric matrix closest to a given value.

% Use existing sparse matrix to generate symmetric one.
load('west0479.mat')
W = west0479;
S = W * W';

%' Extract details of sparse matrix S.
[irow,icol,a] = find(S);
n   = int64(size(S,1));
nnz = int64(size(irow,1));

% zero out strictly upper triangle
for i = 1 : nnz 
  if (icol(i)>irow(i));
     a(i) = 0.0;
  end
end

% Use f11za to remove zero entries.
irow = int64(irow);
icol = int64(icol);
dup  = 'R';
zero = 'R';
[nnz, a, icol, irow, icolzp, ifail] = ...
  f11za(...
        n, nnz, a, icol, irow, dup, zero);

% Evaluate nev eigenvalues (w) closest to sigma.
nev   = int64(20);
ncv   = int64(60);
sigma = 50000;

[nconv, w, v, resid, user, ifail] = ...
    f02fk(...
          n, a, irow, icol, nev, ncv, sigma, 'f02fkz', @option);

fprintf('\n The %d Ritz values closest to %13.5e are:\n', nconv, sigma);

disp(w(1:nconv));

fig1 = figure;
plot(w(1:nconv),'k+');
title('Eigenvalues of WW^T (W=West0479) closest to 50000');
xlabel('index');
ylabel('eigenvalue');


function [istat, user] = monit(ncv, niter, nconv, w, rzest, istat, user);
  istat = int64(0);

function [icomm, comm, istat, user] = option(icomm, comm, istat, user);
    [icomm, comm, ifail] = f12fd('Shifted Inverse', icomm, comm);
    istat = ifail;
f02fk example results


 The 20 Ritz values closest to   5.00000e+04 are:
   1.0e+04 *

    2.5904
    2.8321
    2.9565
    2.9806
    2.9806
    2.9865
    3.1759
    3.3510
    3.4773
    3.6629
    3.6940
    4.0007
    4.0008
    4.9172
    5.3954
    5.6332
    6.0003
    6.2267
    6.8829
    7.2555

f02fk_fig1.png

Optional Parameters

Internally nag_eigen_real_symm_sparse_arnoldi (f02fk) calls functions from the suite nag_sparseig_real_symm_init (f12fa), nag_sparseig_real_symm_iter (f12fb), nag_sparseig_real_symm_proc (f12fc), nag_sparseig_real_symm_option (f12fd) and nag_sparseig_real_symm_monit (f12fe). Several optional parameters for these computational functions define choices in the problem specification or the algorithm logic. In order to reduce the number of formal arguments of nag_eigen_real_symm_sparse_arnoldi (f02fk) these optional parameters are also used here and have associated default values that are usually appropriate. Therefore, you need only specify those optional parameters whose values are to be different from their default values.
Optional parameters may be specified via the user-supplied function option in the call to nag_eigen_real_symm_sparse_arnoldi (f02fk). option must be coded such that one call to nag_sparseig_real_symm_option (f12fd) is necessary to set each optional parameter. All optional parameters you do not specify are set to their default values.
The remainder of this section can be skipped if you wish to use the default values for all optional parameters.
The following is a list of the optional parameters available. A full description of each optional parameter is provided in Description of the s.

Description of the Optional Parameters

For each option, we give a summary line, a description of the optional parameter and details of constraints.
The summary line contains:
Keywords and character values are case and white space insensitive.
Advisory  i
Default =  the value returned by nag_file_set_unit_advisory (x04ab)
If the optional parameter List is set then optional parameter specifications are listed in a List file by setting the option to a file identification (unit) number associated with Advisory messages (see nag_file_set_unit_advisory (x04ab) and nag_file_open (x04ac)).
Defaults  
This special keyword may be used to reset all optional parameters to their default values.
Iteration Limit  i
Default = 300  
The limit on the number of Lanczos iterations that can be performed before nag_sparseig_real_symm_iter (f12fb) exits. If not all requested eigenvalues have converged to within Tolerance and the number of Lanczos iterations has reached this limit then nag_sparseig_real_symm_iter (f12fb) exits with an error; nag_sparseig_real_symm_proc (f12fc) can still be called subsequently to return the number of converged eigenvalues, the converged eigenvalues and, if requested, the corresponding eigenvectors.
Largest Magnitude  
Default
Both Ends  
Largest Algebraic  
Smallest Algebraic  
Smallest Magnitude  
The Lanczos iterative method converges on a number of eigenvalues with given properties. The default is for nag_sparseig_real_symm_iter (f12fb) to compute the eigenvalues of largest magnitude using Largest Magnitude. Alternatively, eigenvalues may be chosen which have Largest Algebraic part, Smallest Magnitude, or Smallest Algebraic part; or eigenvalues which are from Both Ends of the algebraic spectrum.
Nolist  
Default
List  
Normally each optional parameter specification is not listed as it is supplied. This behaviour can be changed using the List and Nolist options.
Monitoring  i
Default = -1
If i>0, monitoring information is output to channel number i during the solution of each problem; this may be the same as the Advisory channel number. The type of information produced is dependent on the value of Print Level, see the description of the optional parameter Print Level for details of the information produced. Please see nag_file_open (x04ac) to associate a file with a given channel number.
Print Level  i
Default = 0
This controls the amount of printing produced by nag_eigen_real_symm_sparse_arnoldi (f02fk) as follows.
=0 No output except error messages. If you want to suppress all output, set Print Level=0.
>0 The set of selected options.
=2 Problem and timing statistics on final exit from nag_sparseig_real_symm_iter (f12fb).
5 A single line of summary output at each Lanczos iteration.
10 If Monitoring is set, then at each iteration, the length and additional steps of the current Lanczos factorization and the number of converged Ritz values; during re-orthogonalization, the norm of initial/restarted starting vector; on a final Lanczos iteration, the number of update iterations taken, the number of converged eigenvalues, the converged eigenvalues and their Ritz estimates.
20 Problem and timing statistics on final exit from nag_sparseig_real_symm_iter (f12fb). If Monitoring>0, Monitoring is set, then at each iteration, the number of shifts being applied, the eigenvalues and estimates of the symmetric tridiagonal matrix H, the size of the Lanczos basis, the wanted Ritz values and associated Ritz estimates and the shifts applied; vector norms prior to and following re-orthogonalization.
30 If Monitoring>0, Monitoring is set, then on final iteration, the norm of the residual; when computing the Schur form, the eigenvalues and Ritz estimates both before and after sorting; for each iteration, the norm of residual for compressed factorization and the symmetric tridiagonal matrix H; during re-orthogonalization, the initial/restarted starting vector; during the Lanczos iteration loop, a restart is flagged and the number of the residual requiring iterative refinement; while applying shifts, some indices.
40 If Monitoring>0, Monitoring is set, then during the Lanczos iteration loop, the Lanczos vector number and norm of the current residual; while applying shifts, key measures of progress and the order of H; while computing eigenvalues of H, the last rows of the Schur and eigenvector matrices; when computing implicit shifts, the eigenvalues and Ritz estimates of H.
50 If Monitoring is set, then during Lanczos iteration loop: norms of key components and the active column of H, norms of residuals during iterative refinement, the final symmetric tridiagonal matrix H; while applying shifts: number of shifts, shift values, block indices, updated tridiagonal matrix H; while computing eigenvalues of H: the diagonals of H, the computed eigenvalues and Ritz estimates.
Note that setting Print Level30 can result in very lengthy Monitoring output.
Regular  
Default
Regular Inverse  
Shifted Inverse  
These options define the computational mode which in turn defines the form of operation OPx to be performed.
Regular OP=A
Shifted Inverse OP=A-σI-1 where σ is real
Regular Inverse OP=A-1
Tolerance  r
Default = ε  
An approximate eigenvalue has deemed to have converged when the corresponding Ritz estimate is within Tolerance relative to the magnitude of the eigenvalue.

PDF version (NAG web site, 64-bit version, 64-bit version)
Chapter Contents
Chapter Introduction
NAG Toolbox

© The Numerical Algorithms Group Ltd, Oxford, UK. 2009–2015