Adaptive Particle Swarm Optimization (APSO) for multimodal function optimization

This research paper presents a new evolutionary optimization model based on the particle swarm optimization (PSO) algorithm that incorporates the flocking behavior of a spider. The search space is divided into several segments like the net of a spider. The social information sharing among the swarms are made strong and adaptive. The main focus is on the fitness of the swarms adjusting to the learning factors of the PSO. The traditional Particle Swarm Optimization algorithms converges rapidly during the initial stage of a search, but in course of time becomes steady considerably and can get trapped in a local optima. On the other hand in the proposed model the swarms are provided with the intelligence of a spider which enables them to avoid premature convergence and also help them to escape from local optima. The proposed approaches have been validated using a series of benchmark test functions with high dimensions. Comparative analysis with the traditional PSO algorithm suggests that the new algorithm significantly improves the performance when dealing with multimodal functions.