摘要
粒子群优化算法的搜索性能取决于算法探索和开发能力的平衡,与算法所使用的拓扑结构相关。现有的粒子群拓扑结构不能较好地平衡算法的探索性能和开发能力。为此,依据低配位数、高堆积密度和3D结构等特征,提出一种新的拓扑设计准则。根据此准则,设计一种菱形十二面体的拓扑结构,该拓扑结构由球体按照六方晶格和面心立方结构堆积而成,是具有最大空间利用率的3D最密堆积结构,且拥有较低的平均配位数。实验结果表明,与现有的拓扑结构相比,该拓扑结构搜索到全局最优值的概率较高。
The ability of taking both the exploitation and exploration into account is a key to ensure good performance of the Particle Swarm Optimization（PSO）algorithm. This ability is to a large extent associated with the topological structures used in the algorithm. Most commonly used topologies are usually not quite favorable for assuring this ability,leading to the so-called dilemma of exploitation and exploration. This paper proposes a new design criteria for topologies by introducing such factors as low coordination number,high packing density and 3D structure. According to this rule,a new neighborhood topology for PSO is designed. The new topology,named rhombic dodecahedron,usually is used in crystallology and formed by packing spheres in hexagonal lattice and face-centered cubics,turns out to be of a 3D-close packing with the maximum space utilization with a low average coordination number. Experimental results on benchmark functions show that the proposed topology has a higher probability of finding the global optimum compared with existing topologies.
出处
《计算机工程》
CAS
CSCD
北大核心
2015年第1期200-206,222,共8页
Computer Engineering
关键词
粒子群优化算法
设计准则
配位数
菱形十二面体
密堆积
3D结构
Particle Swarm Optimization（PSO）algorithm
design criteria
coordination number
rhombic dodecahedron
close packing
3D structure
作者简介
马胜蓝（1986-），男，硕士，主研方向：智能计算，数据挖掘；E-mail：msl1121@vipqq．com
叶东毅，教授、博士生导师
杨玲玲，硕士。
