Title: Strategies for Monitoring Network Flows

Speaker: Professor George Michailidis      http://www.stat.lsa.umich.edu/~gmichail/

Date/Time: Monday, January 26, 2009 at 9:00am – 10:00am

Location: CSRI Building, Room 90 (Sandia NM)

Brief Abstract: In this talk we examine several strategies for monitoring network flows. Specifically, the available data may correspond to aggregate measurements on links or to directly sampled measurements. In the former case, an underlying inverse problem needs to be solved in order to make inferences about changes on characteristics of the flows. In the latter case, the available data provide us with direct information about the quantities of interest, albeit at a cost of increased noise. We discuss the technical challenges posed by the inverse problem and ways to solve it. We then develop a number of monitoring strategies, based either on the aggregate data, or the solution of the inverse problem or on sampled data or on their combination. We examine the merits of each approach on a number of simulated and real data sets.

CSRI POC: Scott Mitchell, (505) 845-7594