CPLEX, no solution returned

Solver related questions
Post Reply
Fred
Posts: 266
Joined: 5 years ago

Re: CPLEX, no solution returned

Post by Fred » 1 week ago

Hi,

It would be helpful if you could share the log, not the lst file (see option logoption if you are not sure how to capture the log).
If your model is so challenging that it needs a supercomputer, it might also be a good idea to use up-to-date software, i.e. GAMS 36 instead of GAMS 26.

Not sure what you mean with 500 simulations.

Your chances to get help will increase if you share your code/data to allow other forum users to reproduce what you are doing.

Best,
Fred

Fred
Posts: 266
Joined: 5 years ago

Re: CPLEX, no solution returned

Post by Fred » 1 week ago

Hi,

Cplex finds a very good integer feasible solution with objective value 2.31965e+09 after 77 seconds. At this stage the best bound is 2.31973e+09. That means the absolute gap is 8 and the relative gap is ~ 0.0034% which is why you see a gap of 0.00% in the log. Closing the gap to zero seems to be challenging. In the following 29 hours the absolute gap is reduced from 8 to 6 and the tree size grows to ~ 270 GB which is huge.

Usually, when there is an objective in the range of multiple billions (which might be an indicator for poor scaling btw), there is no point in trying to further reduce an absolute gap which is already so many orders of magnitude smaller than the objective value (8 in your case). With that said, I suggest that you define a reasonable gap of let's say 0.1% as termination criterion by setting optcr=0.001 and accept the almost optimal solution.

I hope this helps!

Fred

Post Reply