Extreme eigenvalues of Erdös-Rényi graphs

Florent Benaych-Georges
Université de Paris V (René Descartes)

In this talk, we shall present some recent works with Charles Bordenave and Antti Knowles giving estimates for the extreme eigenvalues of the centered adjacency matrices of a possibly inhomogenous Erdös-Rényi graph. In the case of the Stochastic Block Model, these works allow to assess the accuracy of the spectral clustering algorithm, whereas in the case of an homogenous Erdös-Rényi graph, they establish a transition between a compact spectrum and a diffuse one, at a threshold which coincides with the connectivity threshold.

Presentation (PDF File)

Back to Workshop III: Random Matrices and Free Probability Theory