Re: eigs under octave

[David Bateman]

address@hidden wrote:
> so, are here any news about eigs() in Octave? Will it be included in v
> 2.9.10? or only in Octave-forge?
> (I need it in a solver for graph partition by Joao Hespanha (grPartition.m),
> which I intend to connect to the OpenOpt)
> BTW Octave (2.9.9) eig() calling sintax is very far from MATLAB one.
> WBR, D
>   
Its still up to me to phone Sorensen and see why he hasn't responded to
the requests for license clarification for ARPACK..

If you want a graph partitioner, look at metis as it seems accepted that
this is the fastest graph partitioning software out there. Metis can be
used in cholmod and soon in umfpack to speed cholesky and LU
factorization and solving in sparse matrices with suitesparse. Making a
binding direct to metis shouldn't be that hard....

Note however, that metis is officially not free software as there are
clauses in its license on asking permission for commercial
distribution.. Therefore by default octave doesn't use metis in most
cases (cf the fedora binary of octave 2.9.9)

D.


-- 
David Bateman                                address@hidden
Motorola Labs - Paris                        +33 1 69 35 48 04 (Ph) 
Parc Les Algorithmes, Commune de St Aubin    +33 6 72 01 06 33 (Mob) 
91193 Gif-Sur-Yvette FRANCE                  +33 1 69 35 77 01 (Fax) 

The information contained in this communication has been classified as: 

[x] General Business Information 
[ ] Motorola Internal Use Only 
[ ] Motorola Confidential Proprietary