基于偏微分方程的增长网络结构分析

为了研究网络的功能,需要首先研究增长网络的拓扑结构,包括网络的度分布和节点度等。当网络规模足够大时,将网络节点的度看作连续变量,根据网络演化过程中所满足的马尔科夫性,建立网络节点数量的变化方程,从而化简变形得到基于一阶双曲方程的增长网络模型。求解得到了兼具优先和随机2种连接机制的网络度分布P(k)和节点度kt0(t),同时也发现了节点度函数与双曲方程特征线之间的关系。根据网络的演化机制,通过对该增长网络模型进行随机模拟,验证了度分布与节点度理论结果的正确性。将网络的度分布计算转化为偏微分方程求解问题,将节点度的变化视为偏微分方程的特征线,将偏微分方程应用于增长网络的建模中,从而可以解析地对网络结构进行分析。

Structure analysis of growing network based on partial differential equations

The topological structure is one of the most important contents in the complex network research. Therein the node degree and the degree distribution are the most basic characteristic quantities to describe topological structure. In order to calculate the degree distribution, first of all, the node degree is considered as a continuous variable. Then, according to the Markov Property of growing network, the cumulative distribution function''s evolution equation with time can be obtained. Finally, the partial differential equation (PDE) model can be established through distortion processing. Taking the growing network with preferential and random attachment mechanism as an example, the PDE model is obtained. The analytic expression of degree distribution is obtained when this model is solved. Besides, the degree function over time is the same as the characteristic line of PDE. At last, the model is simulated. This PDE method of changing the degree distribution calculation into problem of solving PDE makes the structure analysis more accurate.