A0724
Title: Dynamic clustering for heterophilic stochastic block models with time-varying node memberships
Authors: Kevin Lin - University of Washington (United States) [presenting]
Jing Lei - Carnegie Mellon University (United States)
Abstract: A time-ordered sequence of networks is considered, stemming from stochastic block models where nodes gradually change memberships over time and no network at any single time point contains sufficient signal strength to recover its community structure. To estimate the time-varying community structure, KD-SoS (kernel debiased sum-of-square) is developed, a method performing spectral clustering after a debiased sum-of-squared aggregation of adjacency matrices. The theory demonstrates via a novel bias-variance decomposition that KD-SoS achieves consistent community detection of each network even when heterophilic networks do not require smoothness in the time-varying dynamics of between-community connectivities. The identifiability of aligning community structures across time based on how rapidly nodes change communities is also proven, and a data-adaptive bandwidth tuning procedure is developed for KD-SoS. The utility and advantages of KD-SoS are demonstrated through simulations and a novel analysis of the time-varying dynamics in gene coordination in the human developing brain system.
