Framework for Particle Swarm Optimization with Surrogate Functions
Matthew D. Parno, Kathleen R. Fowler, Thomas Hemker · 2009
Particle swarm optimization (PSO) is a population-based, heuristic minimization technique that is based on social behavior. The method has been shown to perform well on a variety of problems including those with nonconvex, nonsmooth objective functions with multiple local minima. However, the method can be computationally expensive in that a large number of function calls is required to advance the swarm at each optimization iteration. This is a significant drawback when function evaluations depend on output from an off-the-shelf simulation program, which is often the case in engineering applications. To this end, we propose an algorithm which incorporates surrogate functions, which serve as a stand-in for the expensive objective function, within the PSO framework. We present numerical results to show that this hybrid approach can improve algorithmic efficiency. 1