M.S. Thesis Defense - Lexin Liu

Date: 24 Jun, 2008
Time: 2:00 PM
Location: 223 Atanasoff Hall
Topic: A Software System for Identifiability in Causal Bayesian Network
Major Professor(s): Jin Tian


Abstract:

Knowing the cause and effect is important to researchers who are interested in modeling the affects of actions, and Artificial Intelligence researchers are among them. One commonly used method for modeling cause and effect is graphical model. Bayesian Network is a probabilistic graphical model for representing and reasoning uncertain knowledge. It has been used as a fundamental tool and is becoming a more and more important area for research and application in the AI field. A common graphical causal model used by many researchers in AI field is a directed acyclic graph (DAG) with causal interpretation that is known as the Causal Bayesian network (BN). Validating causal models requires that those models impose constraints on the probability distribution that governs the generated data. In this thesis, a software system, which is a set of tools to identify causal effects and to find constraints in a causal Bayesian Networks with hidden variables, is developed. The features of the software system are presented in detail and the applications of the software system are discussed.