Model parameters as uniform intervals [a,b]?
OngoingIs it possible to define model parameters as uniform intervals [a,b] and let Gurobi solve the model by choosing a final parameter value from this interval? I saw an example of this in GAMS where the parameters are inputted as uniform(a,b) and the solver returned final values for not only variables but also parameters. In Gurobi, I tried to add parameters into the model as variables with given lower and upper bounds, but this led an unnecessarily complicated non-linear model where I have products of 3 variables and so on. A quick help would be very much appreciated.
-
Hi Nadi,Parameters cannot be defined as variables in Gurobi. To tune parameters on a given model in Gurobi, you can use the Gurobi parameter tuning tool, which can be invoked either by the grbtune command-line tool or tuning APIs.The Gurobi tuning tool currently does not allow the user to define a range of values for a parameter to limit the search over the parameter space. However, the user can provide a list of base settings to be tried first during the tuning process by setting the tuning-related parameter TuneBaseSettings.Best regards,Maliheh0
-
Hi Nadi,
One of my colleagues, Gwyneth Butera, brought to my attention that the model parameters in your question might not refer to the Gurobi parameters and you might be using this term as it is used in the machine learning community. Could you please clarify what model parameters and variables refer to in your question?
Best regards,
Maliheh
0 -
Hello Maliheh,
Thanks for following up.
I mean the coefficients that normally appear in objective function and constraints in product form (e.g., "c", "d" and "e" in a constraint like "c*x1 + d*x2 <= e" ).
Kind regards,
Serhan
0 -
Hi Serhan,
Sorry for the delay.
As you mentioned, you need to define them as decision variables and set their lower and upper bounds to the values in the given interval. Gurobi does not directly support constraints with multilinear terms (a product of three decision variables, for example). You can, however, model such constraints using a series of bilinear terms. Please check the article How do I model multilinear terms in Gurobi? in case you have not seen it.
Best regards,
Maliheh
0
Please sign in to leave a comment.
Comments
4 comments