Parallel Optimization with multistart and fmincon

3 views (last 30 days)
KELVYN BARUC
KELVYN BARUC on 7 Apr 2012
Hello,
I am running an optimization program with fmincon and multistart. I have 4 workers but I don't know which option is better.
  1. Multistart in parallel always and fmincon with parallel always.
  2. Multistart in parallel always and fmincon with parallel never.
Thanks in advance!

Sign in to answer this question.

Answers (3)

Alan Weiss
Alan Weiss on 16 Apr 2012
parfor has the limitation that inner parfor loops are not run in parallel, only the outermost parfor loop is run in parallel. The situation you ask about is documented in http://www.mathworks.com/help/toolbox/gads/bsc7xh9-1.html#bsc7x1l
The short answer is your option 1 is likely to be slightly faster, but in any case both options run pretty much the same, with just MultiStart running in parallel.
Alan
Sargondjani
Sargondjani on 7 Apr 2012
just try?? this is likely to be very problem dependent...
Paul Kerr-Delworth
Paul Kerr-Delworth on 12 Apr 2012
Hi Kelvyn,
There are some guidelines in the Optimization Toolbox documentation on parallel computing with the Optimization Toolboxes. See the following link for more information
http://www.mathworks.com/help/toolbox/optim/ug/briutum.html
Hope this helps!
Best regards,
Paul

Categories

Find more on Particle Swarm in Help Center and File Exchange

Tags

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!