RE: [Help-glpk] Min. problem with reduced costs < 0, but simplex fails to progress

[Joey Rios]

> > Any other suggestions?  It is a little tricky to test too many things
> > as it takes over an hour to get to the output you see above, but I am
> > willing to try anything at this point.
>
> Please make sure that:
>
> a) column(s) added to the current formulation have correct type,
> bounds, and constraint coefficients;

I don't have any branches in the code that adds columns.  They are added the same way each time.  They are always binary variables (double bounded by 0 and 1) and the coeff's are provided by the subproblems.  They are fine for the 16+ subproblem cases (as they all converge to the same optimum value) and seem reasonable for these non-convergent cases.

>
> b) glp_simplex returns 0;
>

I'll keep an eye on this, but thus far seems like it always returns zero.

> c) glp_get_status returns GLP_OPT (note that if glp_simplex detects
> infeasibility or unboundedness, it returns 0; and if it detects
> this on the very first iteration, it performs no pivots).

This looks like a good thing to check for.  I'll run one and check the status.  If all of the decision variables are double bounded, then we shouldn't see unboundedness in the objective, right?  If I see infeasibility, then there is clearly a bug somewhere in my code.  Thanks again for the help.

---

Hotmail® has ever-growing storage! Don’t worry about storage limits. Check it out.