Originally posted by: SystemAdmin
[prubin said:]
[quote author=cplexuser link=topic=1430.msg4031#msg4031 date=1256665539]
While solving a MIP, sometimes there is a performance degradation
Degradation relative to what?
[quote author=cplexuser link=topic=1430.msg4031#msg4031 date=1256665539]
to find optimal solution. Is there a way to identify which bounds and or constraints contributed to the prolonged solve time?
If a constraint never makes contact with the LP hull of the problem, it's useless and just slowing things down (in the sense of making pivots take a bit longer). Other than that, I don't think it's productive to try to point fingers at a particular bound or constraint. The nature of the problem plays a role in solution time, as do the choices of parameter settings, and of course the model formulation. There can be an entirely different formulation, equally valid in the mathematical sense, that produces a faster solution. Finding it is part science, part art and very much part luck.
I think that denser constraint matrices generally slow down the solution process, so it is possible (not guaranteed) that an alternative formulation that replaces the denser constraints with sparse ones [i]may[/i] help, provided that (a) the sparse constraints are at least as tight as the dense ones when integrality is relaxed (so that bounding is not compromised) and (b) you're not replacing a small number of dense constraints with a googleplex of sparse ones.
/Paul
#CPLEXOptimizers#DecisionOptimization