Hello !

> actually, some of these "optional" things here only need an LP solver
> (not a MILP solver), and Sage does have an LP solver, via
> a standard package CVXOPT.
> It would be nice to get rid of these dependencies on optional
> packages.

Well, the only algorithm which I think could be replaced now is the
matching function, if we can ensure that the solution returned by
cvxopt is an integer one. But it would *really* be nice, if we were to
do that, to use cvxopt through the same LP class and not by defining
manually the matrix, etc, etc..

Theoretically, the same thing should be possible for the flow function
too, though not the way it is defined now : there are optional
arguments to this function to let the user add constraints like "a
vertex receives at most 1 of flow", or even directly the integrity of
the flow, as this function is also used (I believe) to find edge-
disjoint paths between two vertices, etc. I do not know what cvxopt
would return in this case.

Do you know how cvxopt compares for Linear Programs with GLPK/CPLEX/
CBC in term of speed ?

Nathann

P.S. : Oh, and let me say again that I am sorry for these broken
docstrings... All these errors are already fixed in some patches
waiting for review in Graph Theory or Numerical, because I corrected
them while working on other things... I should have taken the time to
write independent patches, but well... :-p

-- 
To post to this group, send an email to [email protected]
To unsubscribe from this group, send an email to 
[email protected]
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org

Reply via email to