A particle swarm algorithm for high dimensional, multi-optima problem spaces

The same mechanisms that are so efficient at finding optima may result in a conventional particle swarm optimisation (PSO) algorithm becoming trapped in a local optimum and unable to escape from this to search for further, hopefully better, optima. This problem becomes more significant as the dimensionality of the problem space increases. A new algorithm that uses waves of swarm particles (WoSP) is introduced that allows a swarm to escape from an optimum and forces it to go on exploring. Results are given for a deceptive problem in both 30 and 100 dimensions. The WoSP algorithm performs well on these problems, encouraging the application of WoSP to other multi-optima high dimensionality problems.