Bare bone particle swarm optimization with integration of global and local learning strategies

Abstract

Bare bone particle swarm optimization (BPSO) possesses self-adapting property and uses fewer parameters resulted in simple implementation and free parameter-tuning. Inevitably, it also tends to converges prematurely, especially for problems with multiple extremes. In this paper, a new method combining global and local learning strategy used in traditional particle swarm optimization (PSO) is devised to improve the performance of the bare bone particle swarm optimization. According to the integration, two variants are proposed. Method is simple and the results are fruitful. Tested on a suite of benchmark functions, unimodal and multimodal functions, justifies the feasibility of the strategy. Both solution quality and convergent speed are better than traditional bare bone particle swarm optimizer.

Multi swarm bare bones particle swarm optimization with distribution adaption

“…One of the centers is using local best positions and is considered to have exploration characteristics, while the other is based on the global best position and is used for better exploitation around the best particle. At early iterations, exploration has a bigger weight which gradually decreases in the favor of exploitation [31,32].…”

A Pair-wise Bare Bones Particle Swarm Optimization Algorithm for Nonlinear Functions

“…To increase the accuracy during the optimization process, a bare bone particle swarm optimization with an integration of global and local learning strategies is proposed by Chen 13 . Moreover, Blackwell formulates the dynamic update rule of particle swarm optimization.…”