Convergence happening before reaching the BarConVTol limit
AnsweredWe are observing that in the barrier Linear Problem model Gurobi is mentioning the solution optimal before reaching the BarConVtol limit. If we set the limit to 1e-4 then it says result optimal as soon the gap reached 1e-2. If I set it to 1e-8 then gurobi says that optimal solution is 1e-6. Is there any other parameter which can over write this or any behaviour which can cause this issue?
Also is optimality tolerance is applicable to all algorithms or specific to some algorithm.
-
Hi Vijay,
I have turned this into a ticket.
Cheers,
David1 -
Hello David,
Thank you for turning it into a ticket.
I am able to found a reason for this issue but I just have a additional question, if you can help me with that: Does optimality tolerance is applicable to all algorithms or specific to some algorithm?
Thank you
Vijay
0 -
Hi Vijay,
OptimalityTol is applied to the simplex algorithm (and related).
If you are using barrier to solve the root relaxation then this will apply during the crossover phase (when simplex is applied).
Cheers,
David0 -
Hello David,
Thank you for letting this know.
You can close this ticket, if I will be having some other question I will raise another.
Thank you
Vijay
0
Please sign in to leave a comment.
Comments
4 comments