The stochastic block model is an inhomogeneous random graph used to model networks with community structure. I will discuss the threshold for approximately and exactly detecting communities and the role played by models in statistical mechanics. In particular there exists a threshold below which the block model is contiguous with respect to the Erdos–Renyi random graph and above which it is possible to recover an approximation to the blocks. We relate this threshold to the extremality threshold of the Ising model on a branching process tree and to the spectra of certain random matric Based on joint work with Elchanan Mossel and Joe Neeman.
Allan Sly ( Princeton University)
Date & Time
Wed, 13 February 2019, 15:00 to 16:00
Madhava Lecture Hall, ICTS Campus, Bangalore