Purdue University Numerical Linear Algebra Group

October 8 (part 2), 2015 at Purdue University.

Talk given:

The Fiedler and PageRank vectors, Part 2

Part 2

The Fiedler and PageRank vectors offer loads of information about a graph’s structure. We will sketch a proof that the Fiedler vector can give a near-optimal cut in a graph, then explore how locally-biased versions of these vectors can identify clusters in graphs, and end by modeling the convergence of these vectors with their “solution paths”, which yield insights into each individual node’s place in the greater scheme of things.