Machine Learning Seminar

Nov 2

Wednesday, November 2, 2016

3:30 pm - 4:30 pm
Gross Hall 330


Tyler McCormick

One scale doesn't fit all: Multiresolution social network models:Many existing statistical and machine learning tools for social network analysis focus on a single level of analysis. Methods designed for clustering optimize a global partition of the graph, whereas projection based approaches (e.g. the latent space model in the statistics literature) represent in rich detail the roles of individual community members. In this talk, I will present a series of examples using data from network surveys, social media, and cellphone call metadata to demonstrate that many pertinent questions in sociology and economics span multiple scales of analysis. Then, I will propose a class of network models that represent network structure on multiple scales. These models differentially invest modeling effort within subgraphs of high density, often termed communities, while maintaining a parsimonious structure between said subgraphs. Finally, I will show that our model class is projective, i.e. consistent under sampling and robust to missing data, highlighting an ongoing discussion in the social network modeling literature relating inference paradigms to the size of the observed graph.


Dawn, Ariel