Page 1 of 1

Resource Limit Influences Computation

Posted: Wed Jun 20, 2018 11:56 pm
by NiobiumFire2
Hi everyone,

I'm finding a strange issue with a few MILPs I'm trying to solve.

I run a model with, say, a resource limit of 2700 seconds and obtain a feasible solution X, where the model has not converged. I then increase the resource limit to 7200 s. The model still does not converge, however, reading the output in the console, the best feasible solution at time ~2700 s is vastly inferior to what was found when the resource limit was 2700 s. Additionally, the solver did not even get to the solution obtained previously.

Put another way, the solver seems to be performing differently depending on the resource limit used, in a way that results in inferior solutions at the same length of run time, for longer total run durations.

I am using GAMS 4.8.5, with solver CPLEX. I use "option reslim = 2700;" to set the resource limit.

Any help would be greatly appreciated!
NiobiumFire