[PAST EVENT] The Structure and Interpretation of Graph Spectral Densities
Speaker: David Bindel, Department of Computer Science, Cornell University
Title: The Structure and Interpretation of Graph Spectral Densities
In this talk, we report ongoing work on the analysis of graphs via global summaries of the eigenvalue distributions and eigenvector behavior. Our approach is drawn from the condensed matter physics literature, where the idea of local and global densities of states is often used to understand the electronic structure of systems, and we describe how these densities play a common role in such seemingly disparate topics as spectral geometry, condensed matter physics, and the study of centrality measures in graphs. We then discuss how structural motifs manifest in the spectrum, give fast algorithms to estimate spectral densities, and conclude with a discussion of some of our current research directions in applying these tools to the analysis of large-scale graphs.
David Bindel is an associate professor in the Department of Computer Science at Cornell University. He received his Ph.D. at UC Berkeley in 2006 and after an NSF postdoctoral fellowship at New York University, he joined Cornel in 2009. He has worked on a range of topics, including high performance scientific computing, graph analytics, applied numerical linear algebra, numerical optimization, data analysis, and tools for electrical power grids and for micro-electro-mechanical systems (MEMS). He has received several prestigious awards, including the Alston S. Householder award, the SIAG/LA Prize, and the ASPLOS most influential paper award.