Dresden 2011 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
SOE: Fachverband Physik sozio-ökonomischer Systeme
SOE 14: Networks: From Topology to Dynamics II (with BP, DY)
SOE 14.6: Talk
Wednesday, March 16, 2011, 12:15–12:30, GÖR 226
Are motifs a myth? — •Jörg Reichardt1, Roberto Alamino2, and David Saad2 — 1Complexity Sciences Center, UC Davis and Würzburg University — 2Aston University, Birmingham
Small subgraphs, called network motifs, have received considerable attention in network research over the last years and are suggested as simple building blocks of complex networks. Motifs are attributed functional significance due to their strong over- or underrepresentation when compared to random null models. However, the link randomized null models used for such comparisons generally match the observed networks only in terms of their microscopic structure, destroying all mesoscopic features and hence give biased estimates of the statistical significance of motif counts in real world networks.
We present a generative probabilistic model based on Exponential Random Graphs plus an algorithm to infer model parameters from a given network. This model allows to generate an ensemble of random null models that matches the observed network with respect to both its microscopic and mesoscopic structural features.
We show that such random null models may result in a much more conservative estimation of the statistical significance of motif counts in real world networks. Further, they maintain the parsimonious explanation of complex networks as a collection of conditionally independent edges.