Comparison of social network structure for particle swarm optimization

Abstract

Swarm Intelligence (SI) originated from the study of colonies, or swarms of social organisms. Studies of the social behavior of organisms in swarms prompted the design of very efficient optimization and clustering algorithms. One of the major techniques in SI is Particle Swarm Optimization (PSO) while it is a technique where several particles (solutions) interacting between each other to find the best solutions. PSO shares many similarities with evolutionary computation techniques such as Genetic Algorithms (GA). GA evolution operators such as crossover and mutation, that chromosomes share information with each other, so the whole population moves like a one group towards an optimal area. Therefore, the various optimization techniques of PSO have been implemented in learning to increase the performance and validate the effectiveness of the social network structure. PSO is a functional procedure by initializing a population of random solutions and searches its member, called particle are initialized by assigning random positions and velocities. The potential particle solutions are then flown through the hyperspace to get the optimum solutions. However, to investigate the efficiency of PSO in optimization problem, a classifier must be incorporated particularly for classification problem. The most common classifier that is normally integrated with PSO is Artificial Neural Network. In this study, PSO is chosen and applied in feedforward neural network to enhance the learning process in terms of convergence rate and classification accuracy.