 |
Center for Nonlinear Studies |
We consider a time series of graphs in which the edges and/or vertices have associated attributes (colors). The problem is to detect a local anomaly (for example a change in the communication pattern and the topics of communications). I will review the previous work and discuss some new theoretical results. Then I will describe some new time series models that generate attributed graphs. If time permits I will discuss some issues of model fitting and model selection.
Host: Joshua Neil, jneil@lanl.gov