removing a constraint changes feasible model to in-feasible !?

[Bob Pay]

Hi, 

I will really appreciate if someone help me to understand something. 

I have this .lp file (attached) which is feasible. I just remove constraint "R6" and then model is in-feasible. I can expect it to be unbounded but definitely not infeasible. What could be a possible reason for this behavior?

thank you.

[des...@gurobi.com]

Hi Bob,

This is due to the poor numerics of your model.

The following warnings from Gurobi are signs that it is facing numerical issues.


Warning: Model contains large bounds
         Consider reformulating model or setting NumericFocus parameter
         to avoid numerical issues.
Warning: unscaled primal violation = 0.075 and residual = 0.075

The reason for the poor numerics are the following variable bounds.

 -infinity <= theta(0) <= 1e+15
 -infinity <= theta(1) <= 1e+15
 -infinity <= theta(2) <= 1e+15
 -infinity <= theta(3) <= 1e+15
 -infinity <= theta(4) <= 1e+15

In general, it is not good to have large bounds such as 1e+15. The best would be to replace this with infinity or specify them as free variables.

To learn more about numerics see:

 http://files.gurobi.com/Numerics.pdf

Amal de Silva
Gurobi Optimization

[Bob Pay]

Amal,

I really appreciate your help. I should change the model accordingly, probably find a better bound on those variables.

Thank you.

Best,