Performance improvement of basic particle swarm optimization algorithm by Lyapunov function modeling of fitness function

Abstract

The paper presents a novel concept of improving the convergence speed and solution quality of particle swarm optimization algorithm by Lyapunov modeling of fitness function. Most of the fitness functions that appear in practice can be transformed into positive definite function by using some minor transformations and shifting the coordinates system in the multidimensional space. The paper demonstrates how these positive definite functions can be transformed to Lyapunov functions and as a consequence how the equation of motion of the particles gets altered to lead to a better convergence speed and superior solution quality compared to those of basic particle swarm optimization algorithm.