Peter S. Dodds



Peter S. Dodds, Duncan J. Watts, and Charles F. Sabel (2003), Ultra-Robust and Scalable Networks Based on Hierarchies, Computing Science and Statistics, 35, I2003Proceedings/DoddsPeter/DoddsPeter.presentation.pdf



Ultra-Robust and Scalable Networks Based on Hierarchies
Peter Dodds, (ISERP, Columbia University), peter.dodds@columbia.edu,
Duncan Watts, (Sociology, Columbia University), djw24@columbia.edu, and
Charles Sabel, (Law School, Columbia University), cfs11@columbia.edu

Abstract

We report on the properties of a theoretical class of communication networks. We construct networks by adding links to an initial hierarchical network based on a two-parameter probability distribution. The networks produced range in form through random, team-based (links are added locally), core-periphery (links are added from the top down), to an intermediate class we term as multiscale networks. We observe certain multiscale networks to be ultra-robust in that they are both resilient to high loads of communication and remain highly connected in the unlikely event of a substantial, targeted loss of nodes. We find these networks require a minimal number of links to be added to achieve their characteristics, on the order of the number of nodes. We also show that ultra-robust networks perform well under a variety of communciation patterns and that their performance scales well with system size. Applications lie in the understanding and potentially the design of robust organizational networks.


Take me back to the main proceedings page.