Impact of Local Search in the Memetic Particle Swarm Optimization

Memetic algorithms have become popular partly due to their high performance in multidimensional and non-differentiable problems and their high exploitation capability due to the implementation of local search algorithms helping the population-based algorithms. They were used to solve several theoretical and practical optimization problems. Despite this, there is still a need to study further these algorithms’ behavior and what makes an algorithm successful in different problems. In this paper, we modeled the interaction networks of four memetic swarm-based algorithms. The interaction networks can capture all the communication between the individuals of the algorithm throughout its execution. These networks are analyzed using several metrics. We successfully demonstrated that the applied methodology could not only show the difference between standard swarm algorithms and memetic algorithms, as already demonstrated in the literature. We can also identify unique characteristics of each memetic variety of the same algorithm. It allows a better understanding of the behavior of convergence, stagnation, and the balance between exploitation and exploration in their various stages of execution.