From help-octave-request at bevo dot che dot wisc dot edu Wed Dec 17 15:17:18 2003
Subject: Re: Octave QR factorization
From: A S Hodel <hodelas at eng dot auburn dot edu>
To: taltman at lbl dot gov
Cc: Bart Vandewoestyne <Bart dot Vandewoestyne at pandora dot be>,    <help-octave@bevo.che.wisc.edu>
Date: Wed, 17 Dec 2003 15:16:48 -0600

When in doubt, consult Golub and Van Loan, MATRIX COMPUTATIONS, 
Johns-Hopkins University Press.
I think it's in its 3rd edition, and can  be purchased in paperback or 
hardcover form.

I've met both of the authors.  Besides being very smart, they're also 
great people.

On Wednesday, December 17, 2003, at 03:00 PM, taltman at lbl dot gov wrote:

> On Dec 17, 2003 at 9:53pm, Bart Vandewoestyne wrote:
>
> Bart.V >taltman at lbl dot gov wrote:
> Bart.V >So this is what you mean i should use instead of Q*R==A:
> Bart.V >
> Bart.V >if ( norm(my_Q*my_R-A)/norm(A) < eps )
> Bart.V >   fprintf('Our calculations are correct!\n');
> Bart.V >else
> Bart.V >   fprintf('Our calculations are wrong!\n');
> Bart.V >end
> Bart.V >
> Bart.V >?
>
> Well, in a particular case, this may work. Not always.
>
> Usually, the '< eps' is written '< tol', where 'tol' is some function
> of the dimensions of the matrices involved, their condition number,
> their rank, and the eps ( not necessarily all of those, but some ). As
> I mentioned before, there's no catch-all way. Every algorithm &
> application will have its own desired precision, and is own precision
> limitations due to round-off error.
>
> Is there a numerical analyst in the house who can be more precise?
>
> Thanks,
>
> ~Tomer
>
>



-------------------------------------------------------------
Octave is freely available under the terms of the GNU GPL.

Octave's home on the web:  http://www.octave.org
How to fund new projects:  http://www.octave.org/funding.html
Subscription information:  http://www.octave.org/archive.html
-------------------------------------------------------------