基于扩散小波的社会网络多尺度分析

摘要

针对社会网络中的同质性关系提出一个基于扩散小波的多尺度分析框架，通过局部相似性度量构造扩散算子，在统一的框架下对社会网络中的结构、内容、用户行为等进行多尺度分析。在合成和真实网络数据上进行实验，与典型算法的对比表明，本算法在无参数的条件下快速收敛并得到更好的结果。%A new multi-scale framework based on diffusion wavelets was proposed to analyze the homogeneous relation-ships, which can be used to conduct multi-scale analysis on structures, contents and user behaviors. The diffusion operator of diffusion wavelet only considers the local similarity in this framework. The experiments on both synthetic and real-world networks show that the proposed algorithm outperforms the typical algorithms in multi-scale analysis without parameters.