Originally posted by: SystemAdmin

[supertcxs said:]

Hi all,

I am so frustrated fixing the bug...

I am using CPLEX 9. The cplex log is

Infeasibility row 'c3107':  0  = -1.
Presolve time =    0.20 sec.

Integer infeasible or unbounded.
Current MIP best bound is infinite.
Solution time =    0.21 sec.  Iterations = 0  Nodes = 0

I have looked into the constraint, but did not find the variable that caused the problem. How can I find more detailed information on why this happens?

Thank you in advance!

Menglin
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

[voydge said:]

can you paste your code?
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

[supertcxs said:]

I have fixed the problem last night. Thank you anyway!


[quote author=voydge link=topic=1307.msg3655#msg3655 date=1248868413]
can you paste your code?

#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

[notdarkyet said:]

If that happens again, try the following things to (hopefully) get a better insight of the cause for the infeasibility:

change problem relaxed
set presolve n
opt

If this shows the infeasibility, you are lucky:
display iis
- and hopefully this gives a small set of constraints and variables, which exhibit the cause.
Unfortunately for some problems this irreducible set is quite large and the steps don't really help. But it's worth a try.

#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

[achterberg said:]

Since CPLEX 10 we have the "conflict refiner" that allows you to identify infeasibilities in your MIP models. This can be of great help for model development and debugging.

#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

[cplexuser said:]

How about the performance of conflict refiner? I have seen conflict refiner gives more precise results than IIS available in CPLEX9. Just eager to hear if they perform (in terms of time) better than IIS..
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

[achterberg said:]

As far as I know, it was not possible with the IIS finder of CPLEX 9 to analyze an infeasible MIP. This was only for continuous problems.

The performance of the conflict refiner is very sensitive to the difficulty of the MIP model. It solves a series of MIPs that are related to the infeasible problem of the user. If the infeasibility of the original model is hard to prove, the conflict refiner has a hard time for analyzing it.
Very often, however, the infeasibility of the original model is already detected in presolve or at the root node. In this situation, the conflict refiner is usually pretty fast, but of course, there is always an exception to the rule...

#CPLEXOptimizers
#DecisionOptimization