Summary
Complex statistical models, built by exploiting conditional independence relationships between parsimonious local submodels, are being increasingly used in many domains. These models can demonstrate great complexity globally, and thus describe and predict complicated situations, yet simple local dependencies facilitate modelling, inference and computation. Such Highly Structured Stochastic Systems (HSSS) are a major tool in image analysis, graphical modelling, expert systems, and much of Bayesian statistics.
Duration
Three years, from 1993-1995
