| Abstract:
Community detection is one of the most fundamental problems in modern
network science. Overlapping community detection, where each node may
belong to several communities, is a particularly challenging and compelling
research direction. The dense mixed-membership stochastic block model
(DMMSB) provides a highly general framework for modeling overlapping
community structures. This problem can be reformulated as reconstructing
the relationships among communities from an observed network. Furthermore,
in the presence of noise, it is essential to develop an optimal algorithm
that achieves error rates matching the theoretical lower bound. This work
will combine the methods in algebra, computer science and statistics.
|
