Particle swarm optimization with dynamic change of inertia weights

Abstract

A novel particle swarm optimization (NPSO) with modified fitness function and dynamic change of inertia weights was proposed for solving complex high-dimensional optimization problems. In this algorithm, both the function value at the searching point and the function change rate at the point were combined into fitness function. This new approach could balance the local searching and the global searching by adopting inertia weight matrix to adaptively and dynamically adjust inertia weights. The convergence degree of every dimension was calculated and the dimension of minimal convergent degree was mutated according to some probability. Experiments on five high-dimension test functions indicate that NPSO can enhance the performance of global searching.