摘要： 为弥补已有算法中单一群体多样性监测方法和早熟停滞的不足，提出L范式多测度群体多样性反馈的PSO算法。利用L范式概念给出位置、速度和自我认知3种群体多样性测度方法，将多测度群体多样性作为粒子群自组织系统的反馈信息，动态调整算法的惯性权值和加速系数，从而实现群体粒子的聚集和发散。基于基准测试函数，给出3种群体多样性的变化特征，比较不同范式不同控制策略下算法的性能。实验结果表明该算法具有更强的全局搜索能力和更高的优化精度。

关键词: 早熟停滞, L范式, 多测度群体多样性, 自组织, 粒子群优化

Abstract: Aiming at the lack of the single perception method of population diversity and premature stagnation, Particle Swarm Optimization(PSO) algorithm based on L norm multi-measures population diversity feedback (PSO-L) is proposed. Position diversity, velocity diversity and self-cognitive diversity using L norm are defined as feedback information of a self-organized particle swarm system, which adjusts the parameters of the proposed algorithm so as to adaptively modify the particles to converge or diverge. The corresponding characteristics of population diversity and the performance of the proposed algorithm on different norms and parameter strategies are performed based on test functions. Moreover, PSO-L, along with other tested algorithms is conducted based on much more benchmark problems. Experimental results show that the proposed method has stronger global search ability and higher accuracy.

Key words: premature stagnation, L norm, multi-measures population diversity, self-organized, Particle Swarm Optimization (PSO)

中图分类号: 

• TP391.06