Community detection for multi-view networks: thresholds and algorithms

Music City Center 

We will discuss community detection in multi-layer and dynamical networks. First, we will describe fundamental information theoretic thresholds for community recovery in these models. Next, we will introduce algorithms based on Approximate Message Passing (AMP) which attain the information theoretic threshold. These AMP algorithms require knowledge of the underlying model parameters. We will use an empirical bayes approach to learn these model parameters---this will yield a fully data driven algorithm for community recovery. Time permitting, we will discuss some applications of this empirical bayes approach to dynamical networks.

Community detection

Approximate Message Passing

Empirical Bayes