 |
Center for Nonlinear Studies |
Networks often exhibit hierarchical organization, where vertices divide into groups that further subdivide into groups of groups, and so on over multiple scales. I will present a general technique for inferring hierarchical structure from observed network data. This technique lets us automatically develop a large-scale summary of a network's structure; it lets us create new benchmark networks which are random,but whose statistical properties are similar to the observed one; and it lets us predict missing connections in partially known-networks with high accuracy, and for more general network structures than competing techniques.
This is joint work with Aaron Clauset and Mark Newman; the paper appeared in Nature 453, pp. 98--101 (2008).
Contact Ron Pistone (pistone@lanl.gov) if you wish to meet with Cris during the day to discuss ideas for collaboration.
Host: Garrett Kenyon, Applied Modern Physics (P-21)