Swarm突现性指标体系及有效性  

Metrics system of swarm emergence and its efficiency

作　　者：吴渝[1] 杨晶晶[1] 陈涌涛[1] 

机构地区：[1]重庆邮电大学网络智能研究所,重庆400065

出　　处：《重庆邮电大学学报（自然科学版）》2011年第6期733-740,共8页Journal of Chongqing University of Posts and Telecommunications(Natural Science Edition)

基　　金：国家自然科学基金(60873079;61040044);重庆市(重点)自然科学基金(2009BA2089;2008BB2241);教育部新世纪优秀人才支持计划~~

摘　　要：针对现有突现性指标的有效性尚不明确,普适性也有待验证的问题。在分析了突现研究常见问题的基础上,对现有定量评价指标体系作了补充,构建了加速度均方差、速度均方差及位置均方差等动力学评价指标,设计了各种实验验证所有指标的评价效果,并举例说明指标的通用性和普适性。实验表明,聚类系数和度分布能够有效表现Swarm系统的网络特性,结构熵在表现稳定突现系统信息变化过程方面比其他指标更有效,内部能量和位置均方差更能够反映系统内部的动力学变化。To solve the problem that the efficiency of current emergence indicators are not clear, and their universality needs to be verified, we add three kinetic indicators into current metrics, including mean square deviation （MSE）of accel-eration, velocity and position. Various experiments are designed to compare the efficiency of all indicators, examples are given to illustrate the versatility and universality of these metrics. Experimental results show that degree distribution and average path length are more effective in representing the network performance in Swarm. Colony entropy, structure entropy and power spectral entropy can only exhibit the evaluation process of stable emergence. Internal energy and position MSE are more efficient in reflecting the dynamics within Swarm.

关 键 词：复杂系统 突现 评价指标 有效性 

分 类 号：TP13[自动化与计算机技术—控制理论与控制工程] TP18[自动化与计算机技术—控制科学与工程]