2:00 pm Wednesday, November 12, 2014
Random Structures: Topology Tomography with Spatial Dependencies by
Darryl Veicht (University of Melbourne, Australia) in RLM 8.136
There has been considerable tomography inference work on measurement networks with a tree topology. Here observations are made, at the leaves of the tree, of `probes' sent down from the root which are copied at each branch point. Inference can be performed based on loss or delay (transit time) information carried by probes, and used in order to recover loss parameters, delay parameters, or the detailed topology, of the tree. In all of these a strong assumption of spatial independence between links in the tree has been made in prior work. I will describe recent work on topology inference, based on loss measurement, which breaks that assumption. In particular I will introduce a new model class for loss with non trivial spatial dependence, the `Jump Independent Models', which are well motivated, and prove that within this class the topology is identifiable. Submitted by
|
|