Summarizing the slices: Sample-based core-periphery classification on complex networks

Core-periphery structure refers to a prevalent property exhibited by many real-world complex networks. The formulation and identification of a network core-periphery structure have been a challenging problem. A classical framework (BE) proposed by Borgatti and Everett defines a core-periphery partition of the network by aligning its nodes with a block model and has been a standard method for this task. This method, however, suffers from high computational costs which make it inapplicable to large networks. Realizing this limitation, we proposed a new framework, which aims to efficiently evaluate core-ness of nodes. Our framework builds a model for core-periphery classification by integrating small samples. The experimental results of six real-world networks shows that our methods can efficiently and effectively identify network core, achieving a running time of less than three hours for a network with about 220, 000 nodes.