Non global solution of bilinear problem
AnsweredI am solving a non-convex MIQCP and depending on the starting conditions. I end up with two widely different objective value, indicating a local solution instead of global solution. First, the problem is olved without objective function (warm-start) and then resolved with a minimization objective function (Locally optimal solution in this case).
Alternatively, when the problem is not solved without objective function (no-warm start). Instead its just solved once and with the same minimization objective as previous one, I get global optimal solution.
I am using Gurobi 9.0.2 with Julia/JuMP
-
The culprit was poor setting of `OptimalityTol` parameter, resetting it to default value solved the issue. Please be extremely careful with parameters affecting the dual problem when using non-convex methods.
0
Please sign in to leave a comment.
Comments
1 comment