Open jdemeyer opened 7 years ago
LP or MILP?
For the research problem LP, but it can work the same way for a MILP.
For LP, try redund
from lrslib
. I don't remember if it guarantees to find all equations. If it doesn't, then there's no way around computing the double description (in sage, setting up the polyhedron already computes that).
For MIP, in general no way around computing the integral hull. We have some methods for that in sage; the best implementation is using normaliz, which also supports the unbounded case. However, there is an interesting software, IPO by Matthias Walter (https://polyhedra-oracles.bitbucket.io/) for cases when computing things like this when computing the full integral hull is out of reach.
Right. My idea was essentially to use this strategy mentioned on the IPO website:
On the other hand, maximizing linear objective functions over these polyhedra (though most often NP-hard) can be done very efficiently for moderate sizes (say n=100), e.g., by mixed-integer programming solvers
His thesis is quite informative: http://matthiaswalter.org/downloads/Dissertation.pdf
Looks like IPO now has a cython interface too. We should get this package into sage.
Unfortunately, IPO depends on SCIP which is not open source.
For a research problem I'm working on, I need to know the set of equalities implied by a LP problem.
Component: numerical
Author: Jeroen Demeyer
Issue created by migration from https://trac.sagemath.org/ticket/23770