I’ve recently found this amazing package, and I wondered what would be a good strategy to parallelize the optimization process under the package.
I have 17 parameters for the optimization, and I’ve been trying to use SAMIN in Optim.jl with multiple-cores and multiple-threading. I would like to compare the performance.
My understanding of the package is that it randomly picks up given a Sobol sequence and find a reasonable regions to be locally searched. If this is correct, is the way to parallelize just launching the same code (e.g., multistart_minimization(multistart_method, local_method, P)) on several distributed cores without adding any configuration?