Codes and Expansions (CodEx) Seminar

Sawyer Robertson (UC San Diego):
Effective Resistance and Conductance for Probability Measures on Graphs

Effective resistance and conductance are concepts borrowed from electrical network theory which are typically defined between vertices of an undirected graph: the resistance distance captures how well connected two vertices are, and effective conductance is its reciprocal and is related to graph boundary value problems. In this talk we explore theory and applications of extensions of these concepts to the probability simplex of the vertices of a graph; in particular we will cover theoretical relationships to optimal transport and minimum cuts on graphs, as well as applications to graph-based semi-supervised learning methods.