Algebra, geometry and combinatorics for network models

Given a large sparse data set, two of the fundamental questions relevant for statistical analysis are: (1) what are model parameters that "best explain the data", in the sense that they make the given data most likely to have been observed under the proposed model?, and (2) can the proposed model even be considered appropriate for the given data?

In traditional statistics, these questions do have an answer: the first is done by computing the maximum likelihood estimators (MLEs), and the second by computing goodness-of-fit statistics. In recent years, data sets coming from social networks, for example, are not amenable to traditional analysis and pose several challenges. In this talk I will describe some of the challenges we face today, and offer examples of techniques from an emerging field of algebraic statistics that can be used to study these problems.

Event Topic

Discrete Applied Math Seminar