Application of Globally Adaptive Inertia Weight PSO to Lennard-Jones Problem

Particle Swarm optimization (PSO), one of the most popular natural computing paradigms, usually requires a large number of fitness evaluations to reach the global optimal solution when applied to some real world problems. Recently a new inertia weight strategy namely Globally Adaptive Inertia Weight (GAIW) has been introduced in PSO which deals with this problem to some extent. In this paper the efficacy of GAIW in PSO has been tested on a computationally challenging global optimization problem namely Lennard-Jones problem. The experiments have been performed for small clusters of 3 to 12 atoms. The computational results thus obtained have been compared with three previously existing inertia weight versions of PSO. The results demonstrate that the use of GAIW in PSO significantly reduces the required number of fitness evaluations and hence the computational time while giving higher success rate.