When is a Swarm Necessary?
Thomas Richer, Tim Blackwell · 2006
This paper compares the performance of particle swarm optimization (PSO) to other optimization algorithms over a continuum of problems. This approach is inspired by state diagrams used in physics. The state space is spanned by the problem parameters, and phases of the diagram are regions where a particular algorithm is more effective. These problems are created by landscape generators. In this report, we generate state diagrams for four optimization algorithms, including PSO, and two types of landscape. The stability of the state diagrams is also tested by varying the number of function evaluations, number of particles (for PSO), and number of dimensions.