My model took more than 14 days to run on i7 core computer with 16GB RAM. And it still does not give an optimal solution yet. What will happen if we stop AIMMS during the middle of running a model when the solution is not optimal yet?
The gap in progress window shows 38.2% for 10 days (see attached picture), then it drops to 38.1% on day-13. From what I understand, the gap is the difference between a best known solution and the best bound. If I set the gap to certain values, which part of the model will be affected? Will all the constraints be satisfied/evaluate? Or will AIMMS only satisfied half of the constraints? For your information, my objective is to maximize the preferences of the lecturer. My preferences parameter has the range values from 1 to 5, where 1 is least preferred while 5 is the most preferred.
Can anyone help me? Thank you in advance!
Best answer by Marcel Hunting
Setting the gap, using the Solvers General option ‘MIP relative optimality tolerance’, has no effect on the model. It will just tell CPLEX to stop the solve after it reached that gap. The solution found by CPLEX will be feasible, so satisfy all constraints. For example, it the ‘MIP relative optimality tolerance’ would have been set at 40% then CPLEX would have finished the solve already, probably with a solution with an objective value of 10190 (or slightly worse).
Note: 16GB RAM is quite low for a model of this size. I would recommend to use at least 32GB RAM. But that will probably not make much difference for the performance, as this seems to be a very difficult problem.