This Series: Complexity & Transitions
About The Series:
Under our 'Complexity & Transitions' series, we host guest lectures given by researchers from a variety of discplines whose research fall under the theme of transitions within complex systems. See our Events for details on our other series Complexity & Society.
Diego Garlaschelli is an Associate Professor at the IMT School of Advanced Studies, Lucca (IT) and at the Lorentz Institute for Theoretical Physics, Leiden University (NL). Since 2011, he has led an interdisciplinary research group with interests including network theory, financial complexity, social dynamics, statistical physics and graph theory. Diego specialises in econophysics and network theory. Within his research group, he combines a theoretical approach, largely based on statistical physics, information theory, discrete mathematics and complexity science, with a data science approach informed by the empirical properties of real-world networks.
Title: Transitions in Complex Networks: From Random Graph Models to Early-Warning Signals
Several real-world complex systems are characterized by a network structure whereby internal components are represented as nodes and interactions are represented as links. Extensive analyses have shown that the topology of real networks is extremely heterogeneous, with different nodes exhibiting very different patterns of connections to the rest of the network. Among the various research lines that this observation has generated, an important challenge is: what is the appropriate "reference shape" for a network, that may serve as a benchmark in order to detect, either statically or dynamically, significant deviations of a real network from the expected reference structure? In this talk, I will discuss a rigorous approach, based on the maximum entropy principle, for generating ensembles of random graphs with given structural properties that appropriately encode the observed heterogeneity of any given real-world network. I will first show the applications of this method for pattern detection in networks and for network reconstruction from partial information, which is a key step in the reliable estimation of systemic risk in financial networks. Then, I will show how maximum-entropy ensembles of graphs can be used as null models for the identitication of transitions in real evolving networks, and consider an application to the identification of early-warning signals of the 2008 crisis in the Dutch interbank network.
Location: Centre for Complex Systems Studies, room 4.16, Minnaert Building, Leuvenlaan 4, De Uithof, Utrecht
The lunch is FREE for all the participants, but please register before Wednesday 4th December.