Monday, October 13, 2014 (3:15 p.m. in Yost 306)

Title: Measuring edge importance for random processes on graphs

Speaker: Deena Schmidt (Postdoctoral Scholar, Case Western Reserve University)

Abstract: Many neural systems can be represented as a stochastic process on a graph.  Recently, Schmandt and Galan introduced the “stochastic shielding approximation” as a fast, accurate method for generating approximate sample paths from a stochastic process in which only a subset of states are observable.  We conducted a rigorous analysis of this stochastic shielding heuristic, deriving a new quantitative measure of the contribution of individual edges in the graph to the accuracy of the approximation.  In this talk, I will discuss our analysis and our extension of this method for a broad class of random graph models and for the Hodgkin-Huxley ion channel model.  I will show how these results shed new light on the contributions of different ion channel transitions to the variability of neural systems.  This approach can be applied to a variety of biological networks and has led to many challenging questions.