Abstract

This extended study presents a hybridization of particle swarm optimization with complex network construction and analysis. A partial population restart is performed in certain moments of the run of the algorithm based on the information obtained from a complex network analysis. The complex network structure represents the communication in the population. We present experimental results of the method alongside with statistical evaluation and discuss future possibilities of this approach. The main goal of the work is not to propose a new highly competitive PSO variant but to present the possibility of using the unconventional tool as an alternative to conventional diversity measures. The main benefit of the network analysis is that it has same-time requirements regardless of the dimension of the problem.