Recommended way of solving multiple similar problems with JuMP

#1

Hi,

I have an application where I have to solve a quadratic problem multiple times, with the only thing changing between problems being the RHS of a constraint. Currently I do this by setting up the problem with the RHS as a JuMP variable, and then looping through the applicable RHS values and fixing the variable as described here in the documentation.

This works fine when I use a local solver, but I’d like to send the problem to Gurobi using their Remote Services. I’ve set everything up, and it works, but the bottle-neck now seems to be having to send the problem and solution back and forth over the network many hundreds of times. Is there a way of setting up a JuMP environment so that I can try to solve all of these sub-problems at once with one optimize call? Or alternatively, what’s the recommended practice for solving many similar problems with only one changing constraint at once?

Thanks in advance for any help

#2

The recommended way is in a loop using JuMP.fix as you are doing.

Gurobi’s cloud services aren’t really intended for this use-case. You could try asking them on their forum https://groups.google.com/forum/#!forum/gurobi