A Directional Scale-based Cosine Similarity Indicator for Adaptive Particle Swarm Optimization

초록

Particle swarm optimization (PSO) is an effective swarm intelligence algorithm, that solves optimization problems by simulating the behaviors of natural swarms like bird flocks. In the original PSO, the control parameters decide the performance of the algorithm. In present studies, researchers often used distance as the assessment criterion, which brings a dimensional ambiguity and ignores the consideration of the searching direction. To solve this issue, a directional scale-based cosine similarity indicator (DSCSI) has been proposed. This improvement enables the controller to achieve rational adjustments based on the unique directional differences in populations. The proposed indicator was applied in adaptive particle swarm optimization (APSO) and tested on CEC2014 with different dimensional information, compared with five state-of-the-art algorithms. The results of the comparison show that the directional scale-based adaptive particle swarm optimization (DSAPSO) has better performance as the dimension increases than the traditional distance-based APSO. © 2025 IEEE.

키워드