Di, 13.06.23. 14.15 Uhr, L 2.201
Abstract:
Model complexity remains a key feature of any proposed data generating mechanism. Measures of complexity can be extended to complex patterns such as signals in time and graphs. We study the well-studied class of exchangeable graphs. Exchangeability for graphs implies a distributional invariance under node permutation and is a suitable default model that can widely be used for network data. For this well-studied class of graphs, we make a choice to quantify model complexity based on the (Shannon) entropy, resulting in graphon entropy. We estimate the entropy of the generating mechanism of a given graph, instead of choosing a specific graph descriptor suitable only for one graph generating mechanism. In this manner, we naturally consider the global properties of a graph and capture its important graph-theoretic and topological properties. Under an increasingly complex set of generating mechanisms, we propose a set of estimators of graphon entropy as measures of complexity for real-world graphs. We determine the large-sample properties of such estimators and discuss their usage for characterizing evolving real-world graphs.
Biosketch:
Sofia C. Olhede is a professor of Statistics at EPF Lausanne. She joined UCL prior to this in 2007, before which she was a senior lecturer of statistics (associate professor) at Imperial College London (2006-2007), a lecturer of statistics (assistant professor) (2002-2006), where she also completed her PhD in 2003 and MSci in 2000. She has held three research fellowships while at UCL: a UK Engineering and Physical Sciences Springboard fellowship as well as a five-year Leadership fellowship, and now holds a European Research Council Consolidator fellowship. Sofia has contributed to the study of stochastic processes; time series, random fields and networks. Sofia was also a member of the Royal Society and British Academy Data Governance Working Group, and the Royal Society working group on machine learning. Most recently she was one of 3 commissioners on a law society commission on the usage of algorithms in the justice system.
