Thomas Opfer

A typical mistake in Benders Decomposition is that some variables have non-trivial bounds (different from 0 or infinity) and not taking the reduced cost vector into account when calculating the dua...

Hi Andy, first of all, if you look into literature, you might find different standard forms for LPs. So you should tell us which standard form you are talking about. Anyway, for every form, you mig...

There might be an easier way, but this should work: I1 = [s[0] for s in I1]

Some information about the gap and other MIP basics can be found here. The question which gap is good enough to stop the code cannot be answered in general. A gap of 0 is obviously always good enou...

The default upper bound should be +INF, so you should not have to specify it explicitly. But of course, it is perfectly ok to do so. Best regards,Thomas

You might also try to continue using your current machine and restrict Gurobi to 6 cores. The remaining 2 cores might be sufficient for your other work. The difference between 6 and 8 cores for Gur...

Something like -inf <= GOAL <= inf should be the correct way to do this.

I guess the problem is that lower bounds of variables are 0 if not specified otherwise.

The question was about infeasible LPs. So I guess, we are looking for the unbounded ray of the dual (if the dual is unbounded and not infeasible). Don't we have to use FarkasDual to query it? I don...

I have no real answer for your question, but please keep in mind, that in general for dual simplex, the first feasible solution that you find is already optimal. So it makes no sense to stop dual s...