Modeling large dense LP with jump


I am solve large dense LP (10^5 variables almost all coeff are non-zero) with jump and cplex.
Now i am modeling the problem with the matrix:
@constraint(m, A*x==b)
But it takes some time to create the model.

Is there some tips to optimize the model creation in JuMP ?


JuMP is designed for sparse problems and uses sparse data structures. There will always be some unnecessary overhead for modeling dense problems. You’re welcome to profile the JuMP code to find the bottleneck, but you may be better served by using the linprog function directly from MathProgBase.

For more technical details, see a previous discussion.

1 Like