[Octave-bug-tracker] [bug #52999] Syntax change in eigs

[Marco Caliari]

URL:
  <http://savannah.gnu.org/bugs/?52999>

                 Summary: Syntax change in eigs
                 Project: GNU Octave
            Submitted by: caliari
            Submitted on: Mon 29 Jan 2018 08:45:46 AM UTC
                Category: Octave Function
                Severity: 3 - Normal
                Priority: 5 - Normal
              Item Group: Matlab Compatibility
                  Status: None
             Assigned to: None
         Originator Name: 
        Originator Email: 
             Open/Closed: Open
         Discussion Lock: Any
                 Release: dev
        Operating System: GNU/Linux

    _______________________________________________________

Details:

>From R2017b, the syntax (and the underlying method) in eigs has changed
(although the old syntax is still supported)

https://uk.mathworks.com/help/matlab/ref/eigs.html?s_tid=gn_loc_drop

Fixing the syntax should not be difficult and should require only a
modification of eigs.m. The underlying method is now the Krylov-Schur
algorithm proposed by G. W. Stewart, which shares with ARPACK the Arnoldi
factorization and other ideas. I am aware only of this free toy
implementation

https://github.com/dingxiong/KrylovSchur

Due to the importance of the ARPACK package (used by octave, scilab, scipy, R,
...), I think that the right way to do could be an extension of the ARPACK
modes to include the new algorithm, or a fork to a KSPACK. In this way, all
the nice features already in ARPACK (single/double precision,
symmetric/unsymmetric/complex routines, reverse communication) would be kept.
Implementing from scratch the Krylov-Schur method in octave would be a huuuge
amount of work.




    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/bugs/?52999>

_______________________________________________
  Message sent via/by Savannah
  http://savannah.gnu.org/