Re: Matrix regression of distnce matrices + non negative least square

[David Bateman]

Jaroslav Hajek wrote:
> On Thu, Sep 10, 2009 at 4:57 PM, Corrado <address@hidden> wrote:
>   
> > Dear Jaroslav, Octave list,
> >
> > here I am again ....
> >
> > How do you calculate distance or dissimilarity matrices in Octave?
> >     
>
> For instance, see
> http://octave.sourceforge.net/doc/f/pdist.html
>
>   
> > and how do
> > you treat triangular matrices + diagonal in Octave?  is there a separate
> > package?
> >
> >     
>
> That depends. Diagonal matrices are a special type.
> Triangular/symmetric matrices are typically treated as full, with
> certain functions (such as matrix division) able to specialize. Of
> course, you can also work with vectors of the upper/lower triangular
> elements, or even with "jagged arrays" - cell arrays of vectors. Your
> choice.
>
> There's also a package for triangular matrices in OctaveForge's
> extra/, but I don't think it's much used.
>
> hth
>   
The triangular matrix code is in octave-forge for the sole reason of 
having an example user defined type in C++. See the comments in the docs

http://octave.svn.sourceforge.net/viewvc/octave/trunk/octave-forge/extra/triangular/doc/triangular.texi?revision=HEAD&view=markup

So not surprising the triangular class in octave-forge isn't "used", as 
its use is as an example for others.

D.