 |
Center for Nonlinear Studies |
We start with an application in cyber-security and illustrate how many problems therein can be abstracted to the analysis of a point process network, or a graph with (marked) point processes occurring on every edge. We show how this structure usefully characterises many other modern datasets. Next some recurring and important statistical themes are identified, for example, measuring information flow, network anomaly detection, and big data. Finally we present some new results and methodology for these problems, including testing for dependence between point processes, calibrating and combining Bayesian p-values, or performing distributed Monte Carlo tests.
Host: Melissa Turcotte