C4I Center Seminar Series:
Automated Construction of Bayesian Networks from Qualitative Knowledge

Dr. Edward J. Wright of Haystax Technology presents
“Automated Construction of Bayesian Networks from Qualitative Knowledge”

A Bayesian Network is a powerful way to represent and perform inference in complex problem domains. A challenge that hinders the wider application of Bayesian Networks is the Knowledge Engineering required to generate the structure and parameters of the Bayesian Network. This talk will present an approach to constructing Bayesian Networks from qualitative knowledge extracted from available knowledge sources like government documents. The knowledge is specified in simple English like statements which contain concepts of interest (hypotheses and indicators) and qualitative assessments for the strength of relationships between indicators and the hypotheses that they are evidence for. An automated process constructs the Bayesian Network, which can be immediately used to perform inference. All of the usual advantages of Bayesian Networks apply, including use of the original Bayesian Network as a tool for communication, and as a starting point for refining the domain knowledge through expert elicitation or learning. The technique has been used for a government customer who observed “for the first time we have a computational model of the concepts in this document!”

Dr. Wright is a senior scientist at Haystax Technology, where he has been developing probabilistic modeling solutions to complex, data intensive, inference problems. He is a retired US Army officer with extensive experience in military topography and providing battlespace awareness to commanders. His original background in the traditional statistical tools of the mapping sciences led to an interest in the broader, more powerful statistical techniques of Bayesian Inference. Over the last 20 years he has developed practical applications that use Bayesian Networks to support inference and decision making in fields of military command and control, intelligence data fusion, missile defense, human behavior modeling, geographic reasoning, Counter Improvised Explosive Device operations, and predicting customer queries to a call center. He has a BS in Forest Engineering from SUNY College of Environmental Science and Forestry at Syracuse University, a MS in Geodesy from Ohio State University, and a PhD in Computational Sciences and Informatics from George Mason University.

3:00 pm - 4:00 pm

Engineering Building Room 4705