News Archives

RSS Feed

Introduction to Research Issues in Stochastic Modeling

September 9, 2004

George Luger <>
Department of Computer Science University of New Mexico

Abstract: We introduce techniques for stochastic modeling. We describe, briefly, full Bayesian inferencing. We discuss the assumptions supporting Bayesian Belief Networks and the complexity reductions resulting from using this technology. We give a simple example of a BBN and discuss the representational limitations of the BBN approach. We next talk about the creation of a first-order system for building stochastic models. Dan Pless, as part of his PhD research at UNM, first designed and built this system. We are currently, with support from the US Navy, rebuilding these first-order stochastic modeling tools in Java. We end with a simple example, built by Chayan Chakrabarti and Roshan Rommohan, of our program doing diagnostic reasoning. The example uses a variant of the Hidden Markov Modeling technology to predict failures in a helicopter rotor system.