有倾向性重连产生的反匹配网络

在具有网络结构的系统中度关联属性对于动力学行为具有重要的影响, 所以产生适当度关联网络的方法对于大量网络系统的研究具有重要的作用. 尽管产生正匹配网络的方法已经得到很好的验证, 但是产生反匹配网络的方法还没有被系统的讨论过. 重新连接网络中的边是产生度关联网络的一个常用方法. 这里我们研究使用重连方法产生反匹配无标度网络的有效性. 我们的研究表明, 有倾向的重连可以增强网络的反匹配属性. 但是有倾向重连不能使皮尔森度相关系数下降到-1, 而是存在一个依赖于网络参数的最小值. 我们研究了网络的主要参数对于网络度相关系数的影响, 包括网络尺寸, 网络的连接密度和网络节点的度差异程度. 研究表明在网络尺寸大的情况下和节点度差异性强的情况下, 重连的效果较差. 我们研究了真实Internet网络, 发现模型产生的网络经过重连不能达到真实网络的度关联系数.

Disassortative networks generated by directed rewiring

The degree correlation of nodes is known to considerably affect the network dynamics in systems with a complex network structure. Thus it is necessary to generate degree correlated networks for the study of network systems. The assortatively correlated networks can be generated effectively by rewiring connections in scale-free networks. However, disassortativity in scale-free networks due to rewiring has not been studied systematically.In this paper, we present the effectiveness of generating disassortative scale-free networks by rewiring the already formed structure of connections which are built using the evolving network model. In the rewiring, two randomly selected links are cut and the four ends are connected randomly by two new links. The rewiring will be reserved if the disassortativity changes to the direction we need, otherwise it will be aborted. However, if one or both of the new links already exist in the network or a node is connected to itself, the rewiring step is aborted and two new links are selected. Our result shows that the rewiring method can enhance the disassortativity of scale-free networks. However, it is notable that the disassortativity measured by the Pearson correlation coefficient cannot be tuned to-1 which is believed to be the complete disassortativity. We obtain that the minimum value of the Pearson correlation coefficient depends on the parameters of networks, and we study the effect of network parameters on the degree correlation of the rewired networks, including the network size, the connection density of the network, and the heterogeneity of node degrees in the network. The result suggests that the effect of rewiring process is poorer in networks with higher heterogeneity, large size and sparse density. Another measurement of degree correlation called Kendall-Gibbons' coefficient is also used here, which gives the value of degree correlation independent of the network size. We give the relation of Kendall-Gibbons' coefficient to network sizes in both original scale-free networks and rewired networks. Results show that there is no obvious variance in rewired networks when the network size changes. The Kendall-Gibbons' coefficient also shows that rewiring can effectively enhance the disassortativity of the scale-free network.We also study the effectiveness of rewiring by comparing it with two sets of data of real Internets. We use the evolving network model to generate networks which have the same parameters as the real Internet, including network sizes, connection density and degree distribution exponents. We obtain that the networks generated by rewiring procedure cannot reach the same degree correlation as the real networks. The degree distribution of real networks diverges from the model at the largest degree or the smallest degree, which provides a heuristic explanation for the special degree correlation of real networks. Therefore, the difference at the end of the distribution is not negligible.