Standard Particle Swarm Optimisation 2011 at CEC-2013: A baseline for future PSO improvements

摘要: In this work we benchmark, for the first time, latest Standard Particle Swarm Optimisation algorithm (SPSO-2011) against 28 test functions designed Special Session on Real-Parameter Single Objective at CEC-2013. SPSO-2011 is a major improvement over previous PSO versions, with an adaptive random topology and rotational invariance constituting main advancements. Results showed outstanding performance of family unimodal separable functions, fast convergence to global optimum, while good was observed four rotated multimodal functions. Conversely, weakest all composition problems (i.e. highly complex specially competition) certain general, towards region optimum achieved, requiring less than 10E+03 function evaluations. However, most SPSO2011 limited capability “escape” from sub-optimal regions. Despite limitation, desirable feature its scalable behaviour, which up 50-dimensional problems, i.e. keeping similar across dimensions no need increasing population size. Therefore, it seems advisable that future improvements be focused enhancing algorithm's ability solve non-separable asymmetrical large number local minima second minimum located far true optimum. This effort providing baseline fair comparison improvements.