[论文解读] Fractal growth of complex networks: repulsion between hubs

Chaoming Song, Shlomo Havlin|arXiv (Cornell University)|Jul 8, 2005

Complex Network Analysis Techniques参考文献 7被引用 3

一句话总结

本文提出了一种基于重正化的增长机制，通过节点之间的强有效排斥力解释了复杂网络的分形结构，防止了枢纽节点的聚集。该模型自然生成模块化、自相似的结构，并表明分形无标度网络在针对枢纽节点的定向攻击下，其鲁棒性显著高于非分形的对应结构。

ABSTRACT

The emergence of universal properties such as the scale-free property self-similarity, and modularity, as key features of complex networks raises the fundamental question of the governing growth process according to which these structures evolve. The possibility of a unique growth mechanism for biological and social networks, as well as computers in the Internet, is of interest to the specialist and the laymen alike, as it promises to uncover the origins of collective behavior. Here, we bring the concept of renormalization from critical phenomena as a mechanism for the growth of fractal and non-fractal modular networks. We show that the key principle that gives rise to the fractal architecture of the networks is a strong effective ”repulsion” between the most connected nodes (hubs) on all length scales, i.e. the hubs tend to be very disperse in the network (and not clump together). We show that the renormalization growth naturally explains to the emergence of modules in biological networks, which is crucial in understanding the structure of the biochemical functional classes. More importantly, we find that the self-similar property of networks significantly increases the robustness of such networks against targeted attacks on hubs, as compared to the very vulnerable non fractal scale-free networks.

研究动机与目标

• 识别驱动复杂网络中普遍特性（如无标度行为、自相似性和模块性）的潜在增长机制。
• 研究单一增长过程是否能够解释生物、社交和科技系统中分形与模块化结构的出现。
• 探讨分形架构与非分形无标度拓扑相比，对网络在枢纽节点遭受定向攻击时鲁棒性的影响。
• 确立枢纽排斥作为塑造所有尺度下高度连接节点空间分布的关键原则。

提出的方法

• 将统计物理中的重正化概念应用于多尺度网络增长建模。
• 引入枢纽节点之间的有效排斥力，防止其聚集，确保在网络中实现空间分散。
• 使用迭代粗粒化过程模拟网络演化，保持跨尺度的结构特性。
• 分析所得网络拓扑以评估自相似性、模块性和对定向攻击的鲁棒性。
• 在相同攻击情景下，比较由枢纽排斥生成的分形网络与非分形无标度网络。

实验结果

研究问题

• RQ1 何种增长机制可解释复杂网络中分形与模块化结构的出现？
• RQ2 枢纽排斥如何影响高度连接节点的空间分布与连通模式？
• RQ3 分形架构在多大程度上增强了网络对枢纽节点定向攻击的鲁棒性？
• RQ4 基于重正化的增长是否能自然产生生物与社交系统中观察到的自相似与模块化网络拓扑？

主要发现

• 枢纽节点之间存在强有效排斥力，是驱动分形网络架构形成的关键机制。
• 通过该机制生成的分形网络表现出自然的模块性，与生物网络中观察到的生化功能类别一致。
• 分形网络的自相似结构显著提升了其在枢纽节点遭受定向攻击时的鲁棒性，优于非分形无标度网络。
• 通过重正化实现的枢纽分散可形成更具韧性的网络拓扑，降低对级联故障的脆弱性。