PSO-ELPM: PSO with elite learning, enhanced parameter updating, and exponential mutation operator

Particle Swarm Optimization (PSO) has been widely used to solve optimization problems. Although a large number of PSO variants have been proposed so far, they suffer from the shortcomings of the original PSO. This paper proposes PSO with Elite Learning, enhanced Parameter updating, and exponential Mutation operator (PSO-ELPM) to balance the exploration and exploitation capabilities of PSO. In this algorithm, the best-performing particles in the population, known as the elites, are used as exemplars to guide the optimization process. The elitism scheme helps to discover valuable knowledge about the solution space. The adopted elites are also used to compute self-cognition coefficients of particles. Additionally, the inverse of the cube root function is applied to ensure a smooth distribution of weight among the elites. The final improvement is to apply an exponential mutation operator, which determines the mutation probability per particle based on the current iteration and its history. The comparisons among PSO-ELPM and 10 state-of-the-art PSO variants on the CEC 2017 benchmark functions reveal that the proposed algorithm yields higher accuracy with acceptable time complexity. According to the Wilcoxon rank sum test, PSO-ELPM at-least outperforms the competitive algorithms on six and seven functions in 30 and 50-dimensional problems, respectively.