The INFORMS Maryland Chapter Presents Guest Speaker Dr. Kevin Ward Drummey National Security Agency “A BAYESIAN BELIEF NETWORKS TEASER” Date: Thursday May 24, 2012 Time: 6PM-7:30PM Place: 308 Sentinel Dr. Annapolis Junction, MD 20701 Room: Conference Room C1004A In this presentation, we provide a short tutorial on Bayesian Belief Networks (BBNs) as a teaser for those who may be interested in this topic for specific applications or research. We introduce a BBN as a graphical probabilistic model and emphasize the connection of nodes and links in a Directed Acyclic Graph (DAG) with their respective random variables and statistical dependence relations in a joint probability distribution. We focus on discrete random variables and show with Conditional Probability Tables (CPTs) how Bayes Theorem and the Law of Total Probability may be employed to compute conditional probabilities in a natural way and also “in reverse” for a DAG in a BBN. We discuss both diagnostic and causal reasoning in a BBN as well as the computational complexity of BBN probability propagation in a DAG, which is facilitated with Pearl’s method. We also mention the importance of probability elicitation and indicate how numerical probabilities may be qualitatively elicited from subject matter experts. During our tutorial, we illustrate the concepts of a BBN with NETICA software by employing several small but effective DAG examples, including an interactive example which allows participants to actually construct and use a simple BBN. About the speaker: Dr. Kevin Ward Drummey is currently Technical Director in the Enterprise Operations Research, Modeling and Simulation Group (OR/M&S) at the National Security Agency (NSA) within the Department of Defense. He is also the founder of NSA’s Summer Program for Operations Research Technology (SPORT) and currently serves as SPORT Technical Director in OR/M&S. In addition, Dr. Drummey is Adjunct Associate Professor of Statistics at the University of Maryland Baltimore County as well as an Adjunct Faculty Member of NSA’s Cryptologic School. Dr. Drummey’s expertise is in statistical data analysis and modeling, distribution theory and simulation, data mining applications, computational algorithm development, and Bayesian statistics. He has published over 30 technical papers in these areas, mostly within NSA, and has served externally as a referee on several notable professional journals. Dr. Drummey has presented papers and chaired sessions at both ASA and INFORMS conferences and has also served on related committees and career panels. Dr. Drummey has a B.A. Magna Cum Laude in Mathematics from Towson University and an M.A. in Mathematical Statistics from the University of Maryland College Park. He earned his Ph.D. in Statistics from the University of Maryland Baltimore County; his dissertation topic was Bayesian robust estimation under the direction of Dr. Bimal Kumar Sinha. Dr. Drummey is also a certified Lean Six Sigma Green Belt and holds professional certifications in data mining from SAS Institute and in modeling and simulation from George Mason University. Dr. Drummey is Past President of the Maryland Chapter of the American Statistical Association and is currently serving as Chapter Secretary/Treasurer. He is also the proud owner of a Buick Lucerne with the license tag “OUTLIER.” If you plan on attending, you must RSVP by May 22, 2012 to Brian Sacash via email: sacash_brian@bah.com