Comparing inertia weights of particle swarm optimization in multimodal functions

Abstract

Particle swarm optimization (PSO) is a state-of-the-art algorithm in meta-heuristic optimization study area. It is a swarm based algorithm that mimic fish or bird's behaviors in the nature. Success rate of convergence in an optimization algorithm depends on control balancing between exploration and exploitation. Inertia weight coefficient parameter controls convergence rate of PSO algorithm. In this paper, different inertia weight: constant, random, linear decreasing and global-local best methods are used in CEC 2017 multimodal benchmark functions. Multimodal functions have huge numbers of local optima. Seven multimodal functions are used with 10 and 30 variable dimensions. Obtained result and run time statistics are compared and shown in graphs.