by
Prof.Scott Kirkpatrick(School of Engineering and CS, HUJI)
→
Europe/Stockholm
132:028
132:028
Description
For about ten years, I have been analyzing large random networks using the tool of k-shell pruning, which separates out nodes in a network according to how many paths each has available to communicate further into the interior of the network. This procedure provides a unique and natural definition of a "core" in a random network. Simple random "preferential attachment" models of networks exhibit perfect power law decays of the sizes of successive k-shells. So does the AS-graph of the Internet, over a considerable range, with one interesting exception -- its core. But recent work shows that restricting attention to objects with limited size or extent as nodes of a network, modifies the power law behavior. Describing the Internet in terms of links between major cities changes the straight power laws into continuous curves. And a recent study of call data records in a European country, analyzed in terms of its k-shells, gives curves with very different appearance. We may hope to infer from these the different social networks that form, mediated by local neighborhood interests, municipal interactions, and business links that connect the whole country.
work done with many collaborators, especially
Avishalom Shalit, Sorinn Solomon, Yuval Shavitt, Shai Carmi, Shlomo Havlin, Daqing Li, Manuel Cebrian and Sandy Pentland.
