[SciPy-user] small error problem

Ryan Krauss ryanlists at gmail.com
Mon Oct 31 10:01:33 CST 2005


I can certainly live with the order e-15 error and agree that this is
not worth working on now that new scipy is nearing release.

I don't know if this means anything and it seems impossible, but one
thing I did to investigate the answer of 0.0 that was being displayed
was to ask for '%0.30f'%mynumber and I got 0.0000000000000000000 (30
zeros).

Ryan

On 10/30/05, Travis Oliphant <oliphant at ee.byu.edu> wrote:
> Ryan Krauss wrote:
>
> >Interesting.  I was starting ipython with -p scipy.
> >
> >If I enter
> >scipy.restore_numeric()
> >
> >I get
> >(Pdb) mat[1,1]/mat[1,0]-mat[0,1]/mat[0,0]
> >Out[4]: -3.5527136788005009e-15
> >both before and after I augment the column of zeros (instead of
> >exactly 0.0 before augmenting and -1.303556573795106e-08 after
> >aygmenting if I don't restore numeric).
> >
> >
> This sounds reasonable.  An answer of 0.0 was just how it was being
> printed (you'd have to look at the bit pattern to know for sure).
>
> So, it definitely looks like some precision issue with different types
> in the old alter numeric.  As the approach is obsolete, that problem is
> going to have to just sit there for a while.
>
> -Travis
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.net
> http://www.scipy.net/mailman/listinfo/scipy-user
>



More information about the SciPy-user mailing list