Conference: Quantitative Network Science


Revealing the dynamics of complex systems

{{_Ltalk:R}} Prof. Dr. Alexandre Bovet
Date: 26.11.20   Time: 13.00 - 13.45   Room:

In this talk I will show three examples of how we can investigate complex systems using dynamical processes on networks in order to provide unique insights about their functioning. The first example is an investigation into the influence of disinformation on Twitter during the 2016 US election. We revealed the different mechanisms of news diffusion governing traditional news and disinformation through the use of network science, opinion mining and activity time series analysis. In the second example, I will present our novel method for multi-scale anomaly detection in networks with node attributes that builds on a graph signal processing framework. In this case, a diffusive process on a network is used in order to simultaneously detect anomalies (i.e., nodes that act as bottlenecks in the diffusive process) and their contexts (i.e., group of nodes retaining the diffusive flow, at all scales in the network). Finally, I will present a novel framework based on Continuous Time Random Walks to analyse the dynamics and structure of time-evolving networks. By considering the covariance of the random-walk trajectories, we capture the temporal evolution of the network's mesoscopic structures (or “communities”), without making any assumptions about the stationarity of the process and by respecting the time ordering of events. This framework opens the doors to the development of new methods capable of disentangling the complex patterns, occurring over a wide range of spatio-temporal scales, of biological, social and economic systems.