When: Tuesday, February 5, 2019, 12:30 pm
Where: Informatics West, Room 232
Speaker: Sadamori Kojaku
Multiple Core-Periphery Pairs in Networks
Abstract: Networks with core-periphery structure comprise core and peripheral nodes, where core nodes are densely interconnected with each other while peripheral nodes are sparsely interconnected with each other. Many empirical networks are shown to be composed of a single group of core nodes and a single group of peripheral nodes, each of which plays distinct roles in the underlying systems, e.g., leaders and followers in social networks. Here we challenge this long-standing observation: a network may be better regarded as a collection of multiple cores and peripheries, as is the case for communities in networks. We show that heterogeneous degree distributions alone explain a single core-periphery structure. Based on this result, we present a novel algorithm to find multiple groups of core-periphery structure in networks that are not explained by heterogeneous degree distributions. We illustrate our algorithm using various empirical networks, including a political blog network, an inter-bank network and a maritime transportation network.
Biography: Sadamori Kojaku received his PhD in Information Science in 2015 from Hokkaido University, Japan. He moved to the University of Bristol in U.K. as a postdoctoral scholar. His research deals with network theory and its applications, in particular, developing algorithms applied to social and economic networks.