|Stochastic Systems Group|
Professor of Electrical Engineering
This talk describes a class of discrete-time Markov models, which we term moment-linear stochastic systems (Sandip Roy's doctoral thesis, 2003), structured so that moments and cross-moments of the state variables can be efficiently computed using recursions in time and in (moment) order. Jump-linear systems, Markov-modulated Poisson processes, and infinite-server queues all turn out to be moment-linear, and therefore can be represented and studied in a common setting. Of particular interest are moment-linear systems with network structure, an example of which is the influence model previously defined and studied in our group (Chalee Asavathiratham's doctoral thesis, 2000). The talk will outline approaches to studying dynamics, estimation and control for such systems, and our current efforts (begun in Carlos Gomez-Uribe's Masters thesis, 2003) at making connections to the extensive literature on inference in graphical models.
Problems with this site should be emailed to email@example.com