Detection and Tracking on Dynamic Random Networks - MP 129-08

Under a Phase II effort for group tracking on dynamic networks Metron proposes: (1) to develop an efficient, approximate inference scheme for tracking groups on dynamic networks based on the exact but inefficient scheme developed in Phase I; (2) to use this approximate inference scheme to explore the phenomena that arise in tracking groups within large populations, and to understand the information-theoretic requirements of group tracking; (3) to develop machine learning techniques to dynamically determine from a streaming data source the parameters needed by the group tracking algorithm; and (4) to the adapt the model to handle simulated or operational data of customer interest and produce software to demonstrate the group tracking functionality on this data.