Hi, my model gives MIP status(107): time limit exceeded

Fixed MIP status(1): optimal

Resource limit exceeded.

I cannot understand the reason. Can you help me?

I tried *Option resLim=7200 but I cant solve it.

problem ipek.gms (8.82 KB)

Hi, my model gives MIP status(107): time limit exceeded

Fixed MIP status(1): optimal

Resource limit exceeded.

I cannot understand the reason. Can you help me?

I tried *Option resLim=7200 but I cant solve it.

problem ipek.gms (8.82 KB)

This is a nasty MIP where the LP relaxation does not help at all and the solver needs to enumerate the entire tree to prove optimality. Depending on the search of the tree that can take very very long. Gurobi does a good job and solves this in ~30 minutes on my machine (see attached log). Finding the optimal solution is no problem, but proving that is indeed the optimal solution is hard. Perhaps you can find a better formulation of the problem where the LP relaxation gives some better bound. Good luck.

-Michael

problem ipek.txt (35.4 KB)

Dear Michael, really thank you for your help.

After editing, I was able to solve the model for small size sampling. However, it gives the following error for a slightly larger sample. Is this an error installing from me? How can I fix? Can you help me?

MIP status(108): time limit exceeded, no integer solution

ipek.gms (35.4 KB)

Hi,

You can increase the time limit via option reslim.

With the latest GAMS version the default is huge but in the past the default time limit has been 1000 seconds… So I guess you use an old version and hit that time limit.

Best,

Fred

No change, it’s a difficult model. This time Gurobi also does not prove optimality. Log of Gurobi run is attached. Good luck! -Michael

ipek.log (116 KB)

Fred thank you but I tried it option reslim=7200. There was no change.

In option reslim=28800 CPLEX no have memory

I have old version (23.5), its new version is very expensive:).

Thanks.

İpek

Michael really thank you for your help, I know its difficult model.

In the article for this sample size optimal solution is 318.15 and gap 26.1.

When the article reports a gap of 26.1 then it can be an “optimal” solution. And indeed Gurobi found a better solution of value 314.31 than your “optimal” one of value 318.15. The gap in Gurobi is worse though. -Michael