Re: Saving data loses information?

[ghaverla]

On Sat, 2 Nov 2013 01:49:43 +0200
Jose <address@hidden> wrote:

> I have a particular case of a covariance matrix with a large
> condition number, whose inverse is problematic

I am not an Octave expert.  In general, methods to invert matrices seem
to favour fast methods, it may be that whatever Octave is doing,
doesn't.

But, I think that if you know a problem is going to result in a large
condition number in advance, you should invert the matrix using
singular value decomposition based methods, and you should "edit" the
singular values before completing the inversion of the matrix.  You can
find recipes about how to edit the singular values (I think Numerical
Recipes consider this), but if this is a repetitive problem for you, you
may want to spend time finding an optimal way to edit the singular
values.

Gord