14.01.2013 Views

Global Optimization Toolbox User's Guide - MathWorks

Global Optimization Toolbox User's Guide - MathWorks

Global Optimization Toolbox User's Guide - MathWorks

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

8 Parallel Processing<br />

8-2<br />

Background<br />

In this section...<br />

“Parallel Processing Types in <strong>Global</strong> <strong>Optimization</strong> <strong>Toolbox</strong>” on page 8-2<br />

“How <strong>Toolbox</strong> Functions Distribute Processes” on page 8-3<br />

Parallel Processing Types in <strong>Global</strong> <strong>Optimization</strong><br />

<strong>Toolbox</strong><br />

Parallel processing is an attractive way to speed optimization algorithms. To<br />

use parallel processing, you must have a Parallel Computing <strong>Toolbox</strong> license,<br />

and have a MATLAB worker pool open (matlabpool). For more information,<br />

see“HowtoUseParallelProcessing”onpage8-11.<br />

<strong>Global</strong> <strong>Optimization</strong> <strong>Toolbox</strong> solvers use parallel computing in various ways.<br />

Solver Parallel? Parallel Characteristics<br />

<strong>Global</strong>Search × No parallel functionality. However, fmincon can use<br />

parallel gradient estimation when run in <strong>Global</strong>Search. See<br />

“Using Parallel Computing in <strong>Optimization</strong> <strong>Toolbox</strong>” in the<br />

<strong>Optimization</strong> <strong>Toolbox</strong> documentation.<br />

MultiStart<br />

ga, gamultiobj<br />

Start points distributed to multiple processors. From these<br />

points, local solvers run to completion. For more details, see<br />

“MultiStart” on page 8-5 and “How to Use Parallel Processing”<br />

on page 8-11.<br />

For fmincon, no parallel gradient estimation with parallel<br />

MultiStart.<br />

Population evaluated in parallel, which occurs once per<br />

iteration. For more details, see “Genetic Algorithm” on page 8-8<br />

and “How to Use Parallel Processing” on page 8-11.<br />

No vectorization of fitness or constraint functions.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!