[SciPy-dev] Problem with flapack

Travis Oliphant oliphant.travis at ieee.org
Fri Feb 15 21:08:05 CST 2002


> > thank you for fixing the problem. Here the result from a short test:
> > >>> from scipy import *
> > >>> from scipy import signal as s
> > >>> s.poly(s.roots([1.,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20])
> > >>>)
> >
> > array([  1.+0.00000000e+000j,   2.-1.89030962e-015j,  
> > 3.+3.97294109e-015j, 4.+2.16188134e-015j,   5.-5.21090017e-014j,
> >               6.-1.44872571e-013j,   7.-2.18532601e-013j,
> >               8.-2.36476441e-013j,   9.-2.38065611e-013j,
> >              10.-4.36395620e-013j,  11.-9.58807728e-013j,
> >              12.-1.45392938e-012j,  13.-1.52765658e-012j,
> >              14.-1.21333755e-012j,  15.-6.78134691e-013j,
> >              16.-2.67591593e-013j,  17.-7.35574790e-014j,
> >              18.-8.47677514e-014j,  19.-1.43588852e-013j,
> >              20.-9.54262406e-014j])
> >
> >
> >
> > It works fine. It should be possible to polish the values, in order to
> > get rid of the small imaginary error, in the function poly.
>
> Yeah, I noticed that too.  I thought we already had done this???  Didn't
> Travis O. put this in at one point.  Maybe it was on a different set of
> functions. Anyway, this should be added...
>

It is added, right now the tolerance is set at 1e-13 which is violated by 
this example.  You could set the tolerance lower, but that doesn't seem 
advisable.

Try an example a little smaller (say to 10 instead of 20) and you will see it 
work.

-Travis



More information about the Scipy-dev mailing list