CBC and CLP solver version with JuMP

Specific Domains Optimization (Mathematical)
1

I am running an optimization problem with CBC solver but it is taking forever with around 1100 binary variables. The same problem is solved in about 2 minutes if I reduce the number of variables to 900. I suspect that this could be due to version of the CBC solver available with JuMP is not up to date. The same problem with CLP solver gets solved in a fraction of a second if binary restriction on the variables is relaxed.

I have following questions:
1- Is the CBC solver up to date?
2- The CBC solver gives me the same answer as CLP solver for LP. Is the CBC solver already embedded with CLP solver or do I need to use the CLP solver separately?
3- How can I specify feasibility tolerance for constraints violation? Using the following code results in an error message
Model = Model(Clp.Optimizer, PrimalTolerance =1e-9, DualTolerance =1e-9)

2

Mixed-integer linear programs are hard to solve, and run-times can increase dramatically with small changes in the model.

You want: Solvers · JuMP

model = Model(Clp.Optimizer)
set_optimizer_attribute(model, "PrimalTolerance", 1e-9)
1 Like
3

Thanks.