Principal Investigator George Verghese
When we model event propagation in large stochastic networks, we find that the configurations and status of individual nodes drift over time. In order to draw inferences about the dynamic behavior of the entire system from limited measurements at selected nodes, we are using probabilistic inference methods related to those developed in artificial intelligence and expert systems. Similar approaches may be useful in analyzing genomic or proteomic data by generalizing existing methodologies based on hidden Markov models. The challenge is determining how to interpret indirect measurements to assess the evolving status of a complex but structured dynamic system.