Social Network-based Swarm Optimization Algorithm

We propose a new population-based optimization algorithm, named Social Network-based Swarm Optimization algorithm (SNSO), for solving unconstrained single-objective optimization problems. In SNSO, the population topology, neighborhood structure and individual learning behavior are used to improve the search performance of a swarm. Specifically, a social network model is introduced to adjust the population topology dynamically, so as to change the information flow among different individuals. Based on the new topology, an extended neighborhood strategy is provided to build a neighborhood for each individual. Different form other forms of neighborhoods, the new structure includes some real individuals connected to the current one and some virtual individuals having better fitness in history, which could bring to more useful information to individuals for avoiding invalid attempts. Furthermore, we propose a new learning framework that defines two different position update methods for two types of individuals with the aim of enhancing the diversity and search ability of the swarm. The performance of SNSO is compared with seven other swarm algorithms on twelve well-known benchmark functions. The experimental results show that SNSO has a better performance than the selected algorithms.