On Wednesday, 1 July 2015 19:12:41 UTC+1, Volker Braun wrote:
>
> The fuzzy zero bound for polyhedra over RDF is 10^-6 (not: 10^-10) which
> matches the cdd sources.
>
> oops, OK, right. Could it be that some rough de-homogenisation is to blame?
Degeneracy of input is relative; here the smallest face has area about 0.5%
of the largest one (as can be seen by doing exact computation), this does
not seem
like an extreme in any way, IMHO.
>
> On Wednesday, July 1, 2015 at 5:50:15 PM UTC+2, Dima Pasechnik wrote:
>>
>>
>>
>> On Wednesday, 1 July 2015 14:59:20 UTC+1, Volker Braun wrote:
>>>
>>> cdd does not correctly handle floating point precision issues. It works
>>> for non-degenerate input, for everything else there are no guarantees.
>>>
>>
>> the input in question is randomly generated...
>>
>> Is there any rationale for _is_zero() for cdd to me 10^{-10} ?
>>
>>
>>>
>>>
>>>
>>> On Wednesday, July 1, 2015 at 12:27:09 PM UTC+2, Dima Pasechnik wrote:
>>>>
>>>> While working on http://trac.sagemath.org/ticket/10276, we had trouble
>>>> running Polyhedron with RDF precision.
>>>> An explicit example showing the issue is in
>>>> http://trac.sagemath.org/raw-attachment/ticket/10276/x.sage
>>>>
>>>> What happens is that Polyhedron() finishes normally, but it creates
>>>> something that looks like inconsistent data
>>>> (a closer inspection reveals that one of the 136 facets of this
>>>> polytope does not "contain any vertices", in the
>>>> sense that the corresponding inequality is not satisfied with equality
>>>> (within the meaningful precision)).
>>>> At least this data is not good enough to compute the face lattice of it.
>>>>
>>>> Without a better knowledge of Polyhedron() code it is hard for me to
>>>> decide whether this is a bug.
>>>>
>>>> Any comments?
>>>> Thanks,
>>>> Dima
>>>>
>>>> PS. if I use non-default 'field' backend I also get some weird error on
>>>> this data, but it works fine if I use
>>>> QQ as the base_ring.
>>>>
>>>>
--
You received this message because you are subscribed to the Google Groups
"sage-support" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/sage-support.
For more options, visit https://groups.google.com/d/optout.
