Community Detection and Seriation for Complex Networks

Abstract: An important part of data science is the ability to extract information from large networks. Namely, in many modern applications, the data is naturally available in networked form: for example, "friend", “follows", "like" and “share" relationships in social media. In other contexts, such as ecological analysis or text mining, correlation or co-occurrence networks provide a helpful abstract representation of complex data. The performance of common link mining tasks, such as clustering or computing node similarity, benefits from the adoption of the assumption that the network formation was the result of a stochastic process informed by a latent reality. This latent reality can take the form of (binary or partial) membership in a number of communities, or an embedding in a feature space. Link mining can then be formulated as the task of retrieving the latent reality.

The speakers in the proposed minisymposium are all engaged in research that aims to provide a theoretical basis for this problem. Their focus is on inferring community membership (community detection), or inferring a latent line embedding or ordering (seriation). However, each of the speakers has a distinct approach to the problem. We expect a lively exchange of ideas, and new collaborations, as a result.

Zoom link for the minisymposium: https://udel.zoom.us/j/91959968261 (no password required)

1:00-1:25 pm Estimating Affinity Between Vertices in Complex Networks (abstract, slides)

Jeannette Janssen (Dalhousie University), Mahya Ghandehari

1:30-2:00 pm Regularity Lemmas for Clustering Graphs (abstract, recorded talk)

Fan Chung (UC San Diego)

2:00-2:25 pm Sampling from Sparse Graphs with Overlapping Communities (abstract, recorded talk)

Samantha N. Petti (Georgia Tech), Christian Borgs, Jennifer Chayes, Souvik Dhara, Subhabrata Sen

2:30-3:00 pm Centrality in Dynamic Competition Networks (abstract, slides, recorded talk)

Anthony Bonato (Ryerson University)

1:00-1:25 pm Reconstruction of Linear Embeddings of Graphons (abstract, slides, recorded talk)

Aaron Smith (University of Ottawa), Jeannette Janssen, Amine Natik

1:30-1:55 pm Learning Mixtures of Permutations from Groups of Comparisons (abstract, slides, recorded talk)

Cheng Mao (Georgia Tech), Yihong Wu

2:00-2:25 pm Projective, Sparse, and Learnable Latent Position Network Models (abstract, slides, recorded talk)

Neil A. Spencer (Carnegie Melon University), Cosma Shalizi

2:30-2:55 pm Detectability Limits in Dynamic Networks with Link Persistency (abstract)

Amir Ghasemian (Temple University), Aaron Clauset, Cristopher Moore