#6456: Upgrade cvxopt in sage from 0.9 to 1.1.3
-------------------------------------------------+--------------------------
Reporter: was | Owner: mabshoff
Type: defect | Status: needs_review
Priority: major | Milestone: sage-4.6.1
Component: packages | Keywords:
Author: Harald Schilly, Dmitrii Pasechnik | Upstream: Completely
fixed; Fix reported upstream
Reviewer: | Merged:
Work_issues: licence |
-------------------------------------------------+--------------------------
Comment(by drkirkby):
Hi Dima, a few points:
* You have addressed most of the concerns I had
* The bug Peter found has been fixed.
* The Solaris/GCC bug has been resolved.
* There's now some doc tests testing the GLPK interface.
* I personally don't have the maths knowledge to review this properly.
If I understand this correctly, the default solver gives a result of
6.2499999 and GLPK gives 6.25. These two are very similar, so although
I've got no idea what sort of relative error can be expected of the
different solvers, intuitively it looks like the two solvers are agreeing
with each other.
I've got no idea if the answer is right though! Is there any theoretical
reason for accepting these answers? Or is the answer accepted just because
that what's the computer gives? If it was possible to compute this in
another way (perhaps using Mathematica or something like that), or better
still a theoretical explanation of why it is right, it would give me
personally more confidence. I really dislike numerical results which are
not substantiated in any way.
Of course, we are using two solvers here, but part of the code is common.
Another concern I have is that the doctest might not be too good across
multiple platforms. It would never surprise me, if on another CPU, the
GLPK solver gave something like 6.2499999 instead of 6.25. Making the
doctest accept any value starting with 6.2 would be silly, as that could
allow huge errors to pass. But I don't know how best to handle this. This
seems a weakness in the way we doctest. Assuming the real answer is 6.25,
we really want something that will accept any x such that abs(x-6.25) <
1e-6 or something like that.
I don't know what this does on SPARC ('mark' or 'mark2' on skynet), but
I'd not be too surprised if the answers came out slightly differently. My
Sun you tested on is not a SPARC processor, but an Intel Xeon.
Overall I'm a '''lot''' happier with this than I was a few months ago, but
don't have the maths knowledge to review it properly and have some
concerns about the numerical results for the doctest.
Dave
--
Ticket URL: <http://trac.sagemath.org/sage_trac/ticket/6456#comment:118>
Sage <http://www.sagemath.org>
Sage: Creating a Viable Open Source Alternative to Magma, Maple, Mathematica,
and MATLAB
--
You received this message because you are subscribed to the Google Groups
"sage-trac" group.
To post to this group, send email to [email protected].
To unsubscribe from this group, send email to
[email protected].
For more options, visit this group at
http://groups.google.com/group/sage-trac?hl=en.
