Abstract:
|
The spread of infectious disease in a human community or the proliferation of fake news on social media can be modeled as a randomly growing network. The history of the random growth process is often unobserved but contains important information such as the source of the infection. In this talk, we model an infection network as the union of a preferential attachment tree, which constitutes the signal, and an Erdos--Renyi graph, which can be regarded as the noise. We consider the problem of statistical inference on the latent history using only a single snapshot of the final network. We provide an MCMC algorithm to construct valid Frequentist confidence sets that scales to networks with hundreds of thousands of nodes. We also bound the expected size of the proposed confidence set for the root nodes and show that in many cases, the size of the confidence set does not increase with the size of the observed network.
|