Jeudi 17 Novembre 2011, 11h, salle 25-26/105
Complex networks are a powerful abstraction that fits a variety of phenomena across several scientific fields, including biology, sociology and economy. Analyzing and extracting insights from large complex networks is an ongoing goal of Complexity Science. In this presentation we present a novel approach based on evolutionary computation and genetic programming. Our method relies on using simple computer programs to represent network generative models, and then applying evolutionary search to find the best generators for observed networks. The final goal of this work is to be able to map large complex networks to plausible generators that have an high explanatory power. For this approach to be successful, a few obstacles have to be overcome. One of these is the measure of quality that guides evolutionary search, which has high overlap with another open problem in network theory: how to measure the distance between large networks with arbitrary sizes and topologies. We present our own solution to this problem using centrality metrics and a well known image recognition algorithm.