EVENT DETAILS AND ABSTRACT

Title: The role of subsampling and exchangeability in modelling (sparse) graphs
Speaker: Victor Veitch
Speaker Info: Columbia
Brief Description:
Special Note:
Abstract:

Most popular statistical models for network valued data fall under the remit of the (dense) graphon framework. These models are seemingly very natural---in particular, they are the models satisfying a certain obvious notion of exchangeability. However, they implicitly assume the networks are densely connected---an assumption that rarely holds for real-world networks of practical interest. Motivated by this, recent work has introduced the sparse exchangeable graphs, and the associated graphex framework, as a as a generalization of dense exchangeable graphs and the associated graphon framework. The development of this subject involves the interplay between the statistical modeling of network data, the theory of large graph limits, exchangeability, and network sampling. In this talk, I will focus on explaining the role of exchangeability and network subsampling in deriving and understanding the new class of models. This leads to a number of interesting results, including a natural unification of statistical network modelling with the theory of large (sparse) graph limits, and the equivalence of exchangeability and invariance of the models under a certain subgraph sampling scheme.