基于概率分布的复杂网络结构模块化方法研究

Research on Modular Method of Complex Network Structure Based on Probability Distribution

提出基于概率分布的复杂网络结构模块化方法.复杂网络结构模块化拥有一个专属流程,分为模块变量分配和结构化索引两个步骤.在模块变量分配中,所提方法构建出复杂网络模块概率分布模型,通过吉布斯采样获取模型学习样本.对样本进行监督,以获取模块条件概率函数真实解.用真实解表示模块比重,真实解的数量表示复杂网络结构模块数.在结构化索引中,所提方法利用LDA主题模型确定以上构建的模块化复杂网络是否成立.实验结果表明,所体方法构建出的模块化复杂网络具备较强线性.

The complex network node and a large number of data constantly changing complex network structure of the modular method, previously proposed mostly lack of universality and linear, cannot effectively improve network performance index engine. In this paper, a method of complex network structure based on probability distribution is proposed. The modular structure of complex network has an exclusive process, which is divided into two steps： module variable allocation and structured index In the module variable assignment, the proposed method constructs a probability distribution model of complex network module, and obtains the model learning sample by Gibbs sampling. The sample is monitored to obtain the real solution of the conditional probability function. The number of real solutions is represented by the number of real solutions. In structured index, the proposed method uses the LDA topic model to determine whether or not the modular complex network is established. The experimental results show that the modular complex network constructed by the method has strong linear.