[changeset] generalized eigenvalues

[David Bateman]

Jarkko Kaleva wrote:
> Hi,
>
> I posted this patch a while back, but it went pretty much unnoticed. So here it
> is again. This is my first attempt in octave development, so any
> comments and suggestions
> are much appreciated.
>
> I decided to take a shot at implementing a feature for solving
> generalized eigenvalue problems.
>
> For EIG and fEIG classes the implementation was quite straightforward.
> Added the needed lapack calls, otherwise the functions are very close
> to the basic eigenvalue functions.
>
> DLD-FUNCTIONS/eig.cc needed a bit more work. But I think that there
> wasn't too much increase in complexity. The original structure is
> still intact, except for the matrix type check, which is now done
> before calculating the eigenvalues.
>
> I added test cases for following input cases (single and double value):
>  - Nonsymmetric real, nonsymmetric real
>  - Symmetric real, symmetric and positive definite real
>  - Hermitian complex, hermitian positive definite complex
>  - Non-hermitian complex, non-hermitian complex
> These tests should go through all the additions to the EIG and fEIG
> classes. (Maybe these should be located in EIG.cc and fEIG.cc)
>
> I also added a couple tests to check the input of invalid matrix types.
>
>   
This seems like a good addition to me, and compatible with Matlab.. I 
see no objection to committing it.. Does anyone else?

D.


-- 
David Bateman                                David.Bateman at motorola.com
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