Originally posted by: guojunyu

I have a very large binary problem, which includes at least 10000 binary variables.
it seems that the cplex can not solve it by using the default setting.
I am just wondering is there any good method in solving the binary only problem.

thanks
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

This is a very hard question to answer in general. It really depends on the model that you are trying to solve. If there was a nice uniform way for solving any pure binary program, then we would have already included this in our default settings.

One possibility is to try the automatic tuning tool that is available since CPLEX 11.0. Another option is to submit your model to our support, and they will figure out settings that you can choose to solve your model faster.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: guojunyu

Thanks for your help.
The tune is trying to change rootalg to PRIMAL OR BARRIER, which can not get the solution neither.
I am using cplex 12.1 I think.
I was in this solving process almost 3 weeks. it is a critical solution to our project.

Could you give me linkage to the support? By the way, the lp file is about 30 MB. Can I submit this file
directly to the service support?

Again, thanks.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: EdKlotz

You can send is the files in several ways.

1) If you gzip or winzip the LP file, you can upload it to this forum question.
Just reply to the thread and use the attach files icon.

2) If you can put it somewhere on your local machines where we can ftp it from
you, that will work as well. Just include the instructions on this thread.

3) If neither of those work, I can open a call for you, and once I do that,
I can send you instructions to upload the file to us.
Let us know which approach works best for you.

Ed
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: JimDan

Hello Ed:

I have a general IP problem (4_20_0_1_1.lp - contained in .zip file) for which I have a starting solution produced via a heuristic (6_20_0_1_1.mst - contained in .zip file). From the cplex.log file (contained in .zip file) it appears CPLEX (under default options) has a tough time improving the lower bound even after 12 hours of computing on a 8 GB RAM Intel Duo Core Machine. Do you/ILOG/IBM have any possible suggestions/MIP settings that can help speed up computation?

I have reasons to believe that my heuristic starting solution is fairly good...So, the trick might be to quickly help improve the lower bound rather than spend time trying to find a better integer feasible solution. I have tried some alternative settings without much success. It is possible these settings together are not effective and some other settings may/may not be needed.

Thanks.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

It seems that your model solves much better with CPLEX 12.2 than with CPLEX 12.1. I ran your model (without your MIP start) on a 4 core Intel Xeon 5160 @ 3.00GHz both with CPLEX 12.1 and CPLEX 12.2. After 20 minutes, CPLEX 12.1 was only able to process about 800 nodes and reached primal/dual bounds of 8200.0000 and 7674.7732, respectively. In contrast, CPLEX 12.2 was able to process 7000 nodes in the same time, getting to bounds of 7820.0000 and 7780.0000.

So, it may be useful to upgrade to the newest CPLEX version. However, even CPLEX 12.2 seems to get stuck at the dual bound of 7780. I only waited for a total of 30 minutes and do not know if CPLEX 12.2 is able to close the remaining 0.51% gap in reasonable time.
Tobias
#CPLEXOptimizers
#DecisionOptimization