CMStatistics 2023: Start Registration
View Submission - CMStatistics
B0882
Title: Approximate inference of network diffusion sources by graphical models Authors:  Tianxi Li - University of Minnesota (United States) [presenting]
Abstract: Inferring the source of diffusion processes on social networks is crucial in fields such as epidemiology and agriculture. However, valid statistical inference is only computationally manageable for specific network structures. It is demonstrated that, while likelihood-based inference is theoretically optimal for general networks, it is computationally infeasible. To resolve this, a class of graphical models featuring network-structured dependence is proposed, which provides an effective alternative. These models enable approximate inference of the diffusion source, thus striking a better balance between computational demand and accuracy.