机构地区: 南京财经大学信息工程学院
出 处: 《计算机应用与软件》 2013年第11期168-171,234,共5页
摘 要: 针对粒子群算法在运行后期易出现种群多样性丢失、早熟收敛这一现象,提出一种基于聚集度与向量相似度的改进的算法。首先基于向量的欧氏距离定义了聚集度的概念,用来衡量种群的多样性;然后采用向量的夹角余弦值来衡量粒子的相似度,为粒子的变异提供依据;最后为了跳出局部最优,对粒子实施散离策略。仿真实验结果表明改进算法具有更强的寻优能力、更快的收敛速度,且解的稳定性更好。 In late stage of its operation, standard particle swarm optimisation is easy to lose the population diversity and to have premature convergence. Aiming at this phenomenon, an improved algorithm based on aggregation degree and vector similarity (ADVS-PSO) is presented in this paper. First, we define the concept of the aggregation degree based on Euclidean distance of vectors to measure the diversity of the population. Then, we measure particle similarity through the cosine value of the angle between vectors to provide the evidence of particle variation. Finally, in order to escape from local optima, we implement a discrete strategy on the particles. Simulation results show that the improved algorithm has more powerful optimisation ability, better convergence speed and more stable solutions.
关 键 词: 粒子群优化 多样性 聚集度 向量相似度 散离策略
领 域: [自动化与计算机技术] [自动化与计算机技术]