Tobias Achterberg

Hi Taylor, column generation really is an art. Typically, you cannot  just say "these are too many variables, let's ignore some and price them in later". This approach would be very similar to the ...

SOS1 constraints are handled via branching. If you have a solution that would be feasible without the SOS1 constraint but that violates this SOS1 constraint (i.e., both z and s are non-zero), then ...

Are you using the "Cuts=3" parameter setting? From the source code, this looks like being the only way of getting cuts before the first call to the user callback function. Regards, Tobias  

The condition number "kappa" is a property of a square matrix. So, there is not a single "kappa" for your model, but any subset of the columns of cardinality equal to the number of rows yields one ...

This is weird. Are you saying that for the exact same input (same model data, same parameter settings) and the exact same hardware and operating system, you are seeing two runs where in one run it ...

We use orbital fixing plus orbital probing. There is the "Symmetry" parameter which allows you to use symmetry breaking more aggressively. This basically adjust some limits on how much effort is sp...

General Constraints are internally translated into linear and SOS constraints, depending on the General Constraint type. If you use the same linearization yourself, there should be no structural pe...

It is a bit strange that the *.lp file is written with your names, because the LP file writer is using the exact same method to check variable names as the *.sol file writer. Note also that you sho...

My guess is that you have at least one variable name with some non-standard character (like a space or a Greek letter), and thus Gurobi decides that it cannot produce a *.sol that can already be re...

Sometimes it helps to use the feasRelax model, see https://www.gurobi.com/documentation/8.1/refman/py_model_feasrelaxs.html#pythonmethod:Model.feasRelaxS By calling m.feasRelaxS(0, False, False, Tr...