Stochastic blockmodeling methods for bipartite graphs
Description: Stochastic block models (SBM) are generative models for network structures such as communities, sets of nodes characterized by being more densely connected with one another than with the rest of the network. One of the main advantages of generative models, such as SBM, is that they open the door to inference: if we are given a network G, we can use techniques from statistical inference to estimate the values of ? that best explain or reproduce the observed pattern of connectivity.
The purpose of the project is to use stochastic blockmodeling techniques to detect network communities (clustering) in bipartite graphs. Bipartite graphs are networks with 2 disjoint sets of nodes such that edges only connect nodes from different sets.
In particular, students should (a) understand how the bipartite SBM works  and (b) implement and evaluate the model with some real networks we will provide. Finally, if there are sufficient students working on the project, then they will be asked to propose and test their own model for directed bipartite graphs.
Recommendation: The candidates must have an advanced knowledge of statistical inference and programming skills.
 Yen, Tzu-Chi, and Daniel B. Larremore. "Community Detection in Bipartite Networks with Stochastic Blockmodels." arXiv preprint arXiv:2001.11818 (2020).