Tobias Achterberg

Gurobi does not support branch-and-price (which means to also add columns at local nodes of the search tree). What you can do with Gurobi is so-called "column generation". This means to solve the r...

No, Gurobi does not support bilevel programming. In rare cases, you can translate a bilevel program into a single-level program and then solve this with Gurobi. But in general, bilevel programming ...

In very rare cases it may even help to also use barrier for the local node relaxations. There is of course the very big downside that barrier cannot benefit from the warm start information that you...

Indeed, with our current development version, I still get the wrong behavior for your model: found 5 solutionsS2 usedD1 usedS1 usedD2 usedD1 used But with my fix I get found 4 solutionsS2 usedD1 us...

It is the only way, because Gurobi does not support quadratic range constraints. But note that as soon as you have a quadratic range constraint (which you would turn into an equation by adding a sl...

Seeing the progress is not so easy when completing the solution pool. Obviously, the gap doesn't give you any indication anymore. The only relevant statistics that you can look at are the number of...

Gurobi has only a very rudimentary support for range constraints. Namely, what we do is to directly add an explicit slack variable for the range constraint. Hence, GRBaddrangeconstr() is nothing el...

No, you need to use PoolSearchMode=2 to really find all feasible solutions. With PoolSearchMode=1 there is no guarantee of exhaustiveness. For example, PoolSearchMode=1 would not disable dual preso...

Hi Martin! You can obtain the desired behavior directly through a combination of parameter and attribute settings. What you would do is to do the first optimization with Crossover=0 as usual. Then,...

The reason is parallel execution. It could be that two parallel threads find solutions at the same time. At the synchronization point, we collect these solutions and present them one by one to the ...