DANCE: A Framework for the Distributed Assessment of Network Centralities

We propose a method for the Distributed Assessment of Network CEntrality (DANCE) in complex networks. DANCE attributes to each node a volume-based centrality computed using only localized information, thus not requiring knowledge of the full network topology. We show DANCE is simple, yet efficient, in assessing node centrality in a distributed way. Our proposal also provides a way for locating the most central nodes, again using only the localized information at each node. We also show that the node rankings based on DANCE’s centrality and the traditional closeness centrality correlate very well. This is quite useful given the vast potential applicability of closeness centrality, which is however limited by its high computational costs. We experimentally evaluate DANCE against a state-of-the-art proposal to distributively assess network centrality. Results attest that DANCE achieves similar effectiveness in assessing network centrality, but with a significant reduction in the associated costs for practical applicability. In contrast to previous work, this outcome allows DANCE to be applied to large-scale networks.