摘要
2.25阶和2.5阶网络零模型在与原始网络具有相同的联合度分布的基础上分别具有相同的平均聚类系数和聚类谱。针对如何快速有效地生成2.25阶和2.5阶零模型,基于随机置乱生成零模型的方法,提出一种生成2.25阶、2.5阶零模型的优化算法-d K-目标保持重连算法。该算法改进了Hamiltonian函数,结合模拟退火算法和Metropolis准则,以2阶零模型为起始网络,通过优化迭代,生成2.25阶和2.5阶网络零模型。通过仿真实验,精确计算了真实网络及其对应的2.25阶和2.5阶零模型的聚类系数和聚类谱,从而验证了提出的算法生成零模型的有效性和准确性。同时,仿真实验分析了算法参数的设置对迭代次数的影响,将提出的算法与现有算法就复杂度进行了比较。分析结果表明,所提出的算法在生成2.25阶和2.5阶零模型时迭代次数明显少于其他算法,表明该算法有效降低了计算复杂度。
The 2.25K and 2.5K network null models have respectively the same average clustering coefficient and clustering spectrum as the original network with the same joint degree distribution. To quickly and effectively generate the 2.25K and 2.5K network null mod- els,based on randomized scrambling method, an optimization algorithm of generating 2.25K and 2.5K null models, named dK-target keeping rewiring algorithm,is proposed. Improving the Harniltonian function, combining with the simulated annealing process and Me- tropolis criteron, and setting 2K null model as the original network,it generates 2.25 K and 2.5 K network null models. The clustering co- efficient and clustering spectrum of the real networks and their corresponding 2.25K and 2.5K null models are calculated by simulation, which verifies the effectiveness and accuracy of generated 2.25K and 2.5K null models. Simultaneously, the impact of parameter settings of the proposed algorithm on the number of iterations is analyzed with simulation. After comparison of the proposed algorithm with the others on complexity ,the simulation shows that the proposed algorithm is significantly less than the others on iteration numbers when gen- erating 2.25 K and 2.5 K null models, which means that it can effectively reduce the computational complexity.
出处
《计算机技术与发展》
2018年第1期121-126,共6页
Computer Technology and Development
基金
国家自然科学基金资助项目(61373136
61374180)
江苏省"六大人才高峰"高层次人才项目(RLD201212)
关键词
零模型
聚类系数
聚类谱
模拟退火
null model
clustering coefficient
clustering spectrum
simulated annealing
作者简介
吴睿（1990-），女，硕士研究生，研究方向为模式识别与智能系统;宋玉蓉，教授，博士生导师，研究方向为网络信息传播及其控制、自适应网络建模、仿真与智能优化。
