No, even knowing the problems would probably not help.
To investigate, you should try to keep either the number of variables or the number of constraints constant, there is no reason to assume they are pulling in the same direction.
You may wish to try a more controlable example to investigate. i.e. Set a number of integer variables to have a unique value from a list of values. Obviously you can change the number of variables at will. You may add additional constraints at will such as the variables sum to less than 20.
Assuming you are using mathprog, you should also look at the constraint matrix produced. Especially in more complex models glpk may eliminate or add to the model. You could also shuffle the rows and cols in the constraint matrix. This will leave the problem unchanged, and so take the same time to solve (or not!).
--
Nigel Galloway
On Tue, Sep 18, 2012, at 02:23 AM, esma mehiaoui wrote:
|
Hello everyone,
I am befog in front of the execution times that i get when i run my MILP formulation.
For instance:
When i run the program with 3581 constraints and 1440 variables i get 9.87 sec for the response time.
When i run the program with 5482 constraints and 2080 variables i get 42.91 sec as execution time.
And when i run the program with 7383 constraints and 2720 variables i get 13.89 sec.
Could someone help me to find an explanation ?
Kind regards
Asma
|
_______________________________________________
Help-glpk mailing list