[Pkg-octave-devel] Bug#348480: sample data

[Drew Parsons]

On Sun, 2006-05-21 at 15:44 +0200, Thomas Weber wrote:
> Using your data file, I get:

> octave:2> norm(y-ynew)
> ans = 6.3300
> 
> While Matlab's result is slightly better (norm of difference is ~0.4), I
> don't see the results to differ that much? Oh, and Matlab warns about
> the polynomial being badly conditioned.
> 


Hmm, I'm inclined to say a norm of 0.4 is much better than a norm of
6.3, rather than only slight better.

But anyway, the renormalisation approach I outlined here results in an
octave norm of 0.082, which is certainly much better than 6.3.  The
renormalisation keeps the polynomial from being badly conditioned.

Drew