A Countable State PGM for Tracking Entity Movement
Zajic Tim · 2016
We consider a probabilistic graphical model for the problem of tracking entities moving among a finite set of sites. The observations consist of counts of the number of entities at sites and during movement between sites. A Bayesian approach is adopted and an importance sampling approach taken to obtaining samples from the model. A backtrack-free proposal distribution is considered and an oracle is obtained through the construction of appropriate network flow problems.