Improving Optimization performance with Parallel Computing

Engineers, scientists, and financial analysts frequently use optimization methods to solve computationally expensive problems such as smoothing the large computational meshes used in fluid dynamic simulations, performing image registration, or analyzing high-dimensional financial portfolios. However, computing a solution can take anywhere from hours to days.
 
This article describes two ways to use the inherent parallelism in optimization problems to reduce the time to solution. The first example solves a mathematical problem using the parallel computing option in Optimization Toolbox™, and requires no code modification. The second, a practical engineering optimization problem, requires a single-line change in code. Both examples use Parallel Computing Toolbox™ and MATLAB Distributed Computing Server™ to automate and manage the parallel computing tasks.
 
By Stuart Kozola, The MathWorks
 
This article was published in MATLAB Digest, March 2009