(Matlab) Fails to solve a small MI-SOCP

[Zohar Levi]

I have a small MI-SOCP that Gurobi (7.0.1) fails to solve:

Optimize a model with 134 rows, 161 columns and 677 nonzeros
Model has 29 quadratic constraints
Variable types: 157 continuous, 4 integer (4 binary)

It seems to run the initial iterations infinitely. CPLEX, for example, doesn't seem to have such a problem.

To reproduce the problem, load the attachment and run:

load sol.mat;

gurobi(sol.solverinput.model, sol.solverinput.model.params);

---

On the subject, where in the doc can I find the meaning of Gurobi's output columns, e.g. "Compl"?

[Renan Garcia]

Your model is on the borderline of being infeasible, and the wide range of the matrix coefficients isn't helping barrier. If I set BarHomogeneous=1 and DualReductions=0, Gurobi 7.0.2 concludes as such. You should check the CPLEX solution for violations.

Also, you can find barrier logging guidance at http://www.gurobi.com/documentation/7.0/refman/barrier_logging.html.

--

---
You received this message because you are subscribed to the Google Groups "Gurobi Optimization" group.
To unsubscribe from this group and stop receiving emails from it, send an email to gurobi+un...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
<sol.mat>

[Zohar Levi]

Thanks, BarHomogeneous=1 did the trick, and the logging doc was helpful.

(I think CPLEX detected infeasibility as well, but it was immediate.)