Error in computing irreducible infeasible subsystem
AnsweredHi all
I'm dealing with the infeasibility of a MILP model and tries to use gurobi_iis to recall the IIS.While the model is infeasible,I fail to compute the IIS.Here's the key message:
Explored 1490 nodes (80819 simplex iterations) in 12.39 seconds (10.87 work units) Thread count was 16 (of 16 available processors)
Solution count 0
Model is infeasible
Best objective -, best bound -, gap -
Although the model turns out infeasible,I fail to compute the IIS(in MATLAB based on yalmip):
[model,recovermodel]=export(Cons,Obj,sdpsettings('solver','GUROBI'));
iis=gurobi_iis(model);
with error message:Gurobi error 10015: Cannot compute IIS on a feasible model.
-
Official comment
-
Gurobi Staff
This post is more than three years old. Some information may not be up to date. For current information, please check the Gurobi Documentation or Knowledge Base. If you need more help, please create a new post in the community forum, or try Gurobot, our chatbot interface offering instant, expert-level support. -
-
-
This may be due to bad numerics. The model could be on the brink of infeasibility and (within numeric tolerances) can be both feasible and infeasible. You can try enforcing stricter, that is smaller, tolerances to force Gurobi to find a more accurate solution. You could also set the NumericFocus parameter to a larger value. Finally, you may want to try disabling presolving (Presolve) to avoid any inaccuracies before claiming the infeasibility.
In any case, you should consult our Guidelines for Numerical Issues for further information on the topic.
Best regards,
Matthias0 -
-
Thank you for your comment! I tried on another PC and it started to compute IIS.So maybe the model is on the brink of infeasibility as you said.I'll check the guidelines to reset parameters pertaining to numerics.
0 -
-
Hi Matthias,
I've followed your advice and also revised some constraints.Now the model seem to be able to run but new problems emerge:
Warning: max constraint violation (7.9532e-01) exceeds tolerance
(model may be infeasible or unbounded - try turning presolve off)
I tries to reset the Presovle parameter to 0,and the NumericFocus parameter to 1,2,3.Until the NumericFocus parameter is set to 2,the warning message no longer appears.I notice that the optimal solution remain almost same despite the reset of these parameters.As the model has plenty arows,colums and variables,and the reset of these parameters would lower the efficiency immensly,I'm wondering if it would be better to neglect numeric issues?
Thanks in advance
Suyue Xu
0 -
-
-
Whether or not you can ignore the numerical issues is up to you. You should check the solution and verify that the solution values make sense for the model and the application. Being more careful regarding numerics is usually leading to a worse performance.
Cheers,
Matthias0 -
Post is closed for comments.
Comments
5 comments