Sažetak KiDM 20100531

Izvor: KiWi

(Joint work with Damir Vukičević)

We present a structural analysis of complex networks. More specifically, we speak about communities and hubs in the networks. We examine Estrada’s method for community structure detection and some of its properties. We additionally propose a novel method for fast detection of hubs within a system. The algorithm identifies a set of nodes in the network as most significant, aimed to be the most effective points of distribution for fast, widespread coverage throughout the system. We show that our hubs have in general greater closeness centrality and betweenness centrality than vertices with maximal degree, and have in general higher degree than vertices with greatest closeness centrality and betweenness centrality. As such, they serve as all-purpose network hubs. Several theoretical and real world networks are tested and results are analyzed. We further develop a classification of three types of hubs related to the spread of SIR-type epidemics. We define a network outhub as a node which, if infected first, causes the most extensive spread of a disease throughout the network. An inhub is a node which is most likely to become infected at some point during the spread of the disease through the system. And a network transitive hub is a node which, if vaccinated, most decreases the spread of the disease through the system. We show that on some networks, these three hubs are distinct, indicating the need for more specific identification of network hubs than the general use of vertex degree and connectivity