Re: help in modifying __qp__.cc

[A. Scottedward Hodel]

On Nov 3, 2008, at 10:37 AM, Søren Hauberg wrote:

> man, 03 11 2008 kl. 11:12 -0500, skrev John W. Eaton:
> > After giving this some more thought and looking at the __glpk__.cc
> > file, it might be best to create another level of abstraction that we
> > can add to liboctave.  At the very least, it would probably be good  
> > to
> > have a function like
> >
> >  ColumnVector
> >  glpk (const ColumnVector& c, const Matrix& C, const ColumnVector& d,
> >        const glpk_options& opt);
>
> This isn't really my field, so I can't help out here. But just a small
> question: wouldn't it make sense to call this function something like
> 'solve_linear_program' or 'linear_programming'. That would make it
> easier if we at some point would use a different framework for solving
> these kinds of problems (if one day they chose a non-silly license, we
> just might use http://www.coin-or.org/).
>
> Søren

It's been awhile since I wrote my own interior point method QP code  
(not ready for prime-time), but is there a reason not to use an  
economy size rank revealing QR decomposition [q,r,p] = qr(A,1) at  
start to reduce the dimension of the constraints if they are or can be  
redundant?  The flop cost would be less than one iteration of a QP  
iterative method.

A. Scottedward Hodel address@hidden
http://homepage.mac.com/hodelas/tar