 |
Center for Nonlinear Studies |
Communities play an important role in networks and are relevant to a multitude of scientific fields. However, their structure is not well understood. The social network analysis literature contains a large variety of proposed community structure metrics and algorithms to identify communities, with little consensus on which of these are correct. In this talk, I will begin with an overview of communities in networks, including a discussion of applications and importance to other fields. I will then describe the Community Structural Analysis Framework, a machine learning framework for comparing the structure of communities produced through different methods. Next, I present an application of this framework to a variety of networks, each associated with metadata allowing for the identification of real communities. Using the Community Structural Analysis Framework, one can compare the structure of these real communities to the structure of communities identified by various community detection algorithms. These experiments produce several surprising and valuable insights about the structure of real communities.
Bio:
Sucheta Soundarajan is a postdoc at Rutgers University. She received a PhD in Computer Science from Cornell University in 2013 and a B.S. from the Ohio State University. Her areas of interest include the structure of social networks, communities in social networks, and
interesting applications of social network analysis. Her website is
http://www.cs.rutgers.edu/~ss2078
Host: Aric Hagberg