nested parallel optimization and parfor loops

I would like to run a code having this structure:

Parallel Optimization 1(Particle swarm):

--- Par for loop (length ~ 20):

--- --- Parallel Optimization 2(Particle swarm)

--- calculate a cost function based on the 20 results of the parfor loop and use it in the Parallel Optimization 1

Is it possible to connect to a cluster and make a such complicated nested parallel compuitation?

I am doing this:

Par for loop

Parallel Optimization 2(Particle swarm)

on my computer for now but i am not sure the particle swarms actually works in parallel.

Please tell me if it is doable. Thank you.

Answer 1

[Walter Roberson]

If you specify the UseParallel option as true, then https://www.mathworks.com/help/gads/particle-swarm-options.html#bued1yt-1 calculations will be carried out using parallel processing -- parfor.

Within a parfor worker, parfor (and other parallel requests such as parfeval) are carried out in serial.

It is not possible to parfor within a parfor. (Well, it is possible, but the inner parfor will be executed in serial.)

Note that if you do not specify UseParallel, and instead specify UseVectorized then "Your objective function should accept an M-by-N matrix, where each row represents one particle, and return an M-by-1 vector of objective function values.". Your code might potentially use parfor during handling of that M-by-N matrix.