Michel Soares

MIQCP is a convex optimization problem with integer variables. All convex optimization and integer programming optimization theory still holds and a lot of the technique Gurobi's uses is directly i...

Hi, Indeed there are additional steps involved in equal constraints. However, I cannot see how it would be better to replace it by two inequalities, it should be worse because it increases the numb...

As far as I understand, Aggregate and AggFill will mostly remove variables. These are sometimes referred to as "Substitute implied free variables". Disabling Aggregate increases your model solve ti...

Hi, I am not sure if it would help, but it seems like you do not need Bp[edge], you can use direction_AB and direction_BA instead. It should remove some variables and make your model more dense, wh...

Hi, You can get the presolved model this way: reduced_model = model.presolve() reduced_model.write("reduced_model.lp") You may be able to directly modify the reduced_model directly as you wish, I h...

Hi, Could you write Gurobi's logs? There might be more information there regarding the use of warm start, information about your model and how it is solving your LP.

Hi, The presolve step will usually change your model to another equivalent model that is easier for the solver to find a solution. In this step the number of constraints and variables are usually r...

Hi, The only way I think of doing this would be to disable all pre-solve parameters except PreDepRow and export the reduced model, comparing to the original model. I have never tried this, but it m...

Hi, I am not sure what you mean, but you might want to look into the theory behind branch and cut, which is the algorithm Gurobi uses, and it might get you to your convergence proof.  

Hi Maria, I believe var_i =1 and var_j =1 then dist_ij>= min_dist could be modelled as: dist_ij >= (var_i * var_j  * min_dist) That would make your model quadratic, but it should yield the desired ...