Locally valid user cuts and upper bounds

Hi,

I am solving a mixed integer liner program with big M constraints (coming from linearisation), using the JuMP package in Julia. This big M values correspond to upper bounds of some variables and are highly correlated with the quality of the relaxation. Consequently, I was considering adding user cuts, valid only in subtrees of the branch-and-bound tree, that would use the local bounds of my variables to tighten these constraints. I know that changing the constraint is not possible during the solving but adding cuts should be doable. However, I didn't find a way to access the local bounds of the variables in callbacks.

Is there a way to implement this procedure?

Thanks

Riley Clement

Gurobi Staff

June 22, 2025 14:20

Hi Adrien,

There are a couple of obstacles here. The first is you can't access local bounds, the second is any cut must be globally valid.

- Riley

Adrien Belfer

June 23, 2025 07:51

Hi Riley,

I totally understand the obstacles, but wouldn't it be interesting as a future feature to enable locally valid cuts?

Adrien

June 23, 2025 08:15

Indeed it would be interesting. This feature was added to our backlog some time ago but has not been included in the roadmap yet. Unfortunately there does not look to be a lot of interest from our commercial customers for this feature.

June 23, 2025 08:43

Hi again Riley,

From a commercial point of view, I totally understand the lack of interest. It was more for a research purpose. In any case, thank you for your feedback.

Locally valid user cuts and upper bounds

Comments

Didn't find what you were looking for?