Do topological properties have any impact on the search behavior of population-based stochastic optimization methods like particle swarm optimization (PSO)? Can a PSO algorithm improve efficiency if such properties are tuned adaptively while integrating cellular learning automata (CLA)?

Swarm behavior is behavior that is displayed by a flock of birds, wherein a single bird (also called a particle) utilizes its own historical positions and global best positions to update its own flying velocity and associated movement within the swarm. This approach has often been cited as a reason for the major limitation of getting trapped in local optima.

To overcome this limitation, the paper suggests extending a single swarm model into an interacting multi-swarm model, significantly altering the topological properties. The authors introduce a new attribute, the connectivity degree, into the topological properties. They introduce and deploy a CLA to learn the connectivity degree based on the particle’s historical and social experiences.

There are multiple swarms in the new approach, and each particle is updated based on information from its present parent swarm (where it currently resides) as well as neighborhood swarms. The global best position of each neighborhood swarm is used in the update equation of the particle, offering a superior learning technique for the particle and thus the swarm.

The authors conclude that altering topological features by introducing the multi-swarm concept together with a CLA approach is indeed possible, and that the efficiency of the newly proposed PSO approach could even be superior to previous PSO approaches. The authors compare their novel approach with a few other developed stochastic optimization methods on two sets of benchmark problems, that is, “some classical well-known functions” and CEC2013. The results, which validate the authors’ claims, are included.