|
Abstract:
|
Network-theoretic methods provide a powerful framework to describe complex systems and capture various dependency structures. Spreading processes, i.e. temporal stochastic processes describing propagation through a medium, are of considerable interest to network-based systems analyses. A prominent example is the dissemination of information in social networks. Most work to date studying spreading processes on complex networks has focused on predicting trajectories of spread. Our work considers the reverse problem, the estimation of the origin of a spreading process on a complex network, given snapshot observations of that process at the nodes of the underlying network. Several source estimates have been suggested in different settings. Here, we explore the question of uncertainty quantification for this problem, by assessing the second central moment of a source estimate. For that, we investigate the possibility to utilize different sub-sampling strategies of the spreading process and discuss the results for a handful of scenarios. We illustrate the approach by the identification of effectors in social networks.
|
