Re: Equivalent of LSQR in matlab

[Jordi Gutiérrez Hermoso]

On 13/09/2007, Jean-Baptiste Poullet <address@hidden> wrote:
>(x=A\b is not sufficient given A is large (M>>N) and ill
>conditioned)?

It usually is. Why don't you try it?

You can also try the gls or ols functions.

>I'm also wondering whether the matlab function LSQR (based on
>Conjugate Gradients on the Normal Equations) exists in octave.

The lsqr function you requested is still in our wishlist:

      http://www.gnu.org/software/octave/projects.html
      http://octave.sourceforge.net/doc/L.html#L

But it shouldn't be too hard to code. The conjugate gradient method is
a fairly easy algorithm to implement. Why don't you do it yourself and
then submit your implementation to octave-forge?

- Jordi G. H.