PSO Evolution Based on a Entropy Metric

Abstract

Bioinspired search algorithms are widely used for solving optimization problems. The evolution progress isusially measured by the fitness value of the population fittest element. The search stops when the algorithm reaches a predetermined number of iterations, or when no improvement is achieved after some iterations. Usually, no information, behind the best global objective value, is fed into the algorithm to influence its behavior. In this paper, a entropy metric is proposed to measure the algorithm convergence. Several experiments are carried out using a particle swarm optimization to analyze the metric relevance. Moreover, the proposed metric is used to implement a strategy to prevent premature convergence to suboptimal solutions. The results show that the index is useful for analyzing and improving the algorithm convergence during the evolution.