Grant: Research Tools and Instruments - Category 1 (<$150,000)

Bayesian networks are a means of exploiting conditional independence information in joint probability distributions to facilitate both a compact representation of a joint distribution, and faster inferences from that distribution, without sacrificing the integrity of the original information. This is crucial for probabilistic decision-making in settings where variables interact in complex ways, such as medical decision making. It also facilitates and improves the accuracy of classification, forecasting, and fault diagnosis. Any software that improves the efficiency of decision-making, without sacrificing accuracy, makes deployment on consumer hardware more feasible. One of the major criticisms of using Bayesian networks in application is that their computational demands are often very high, both in terms of runtime and memory. In our proposed research program, we intend to develop streamlined data structures for computing over Bayesian networks at runtime. These data structures will be constructed according to application constraints on resources such as time and memory. Finding the optimal runtime structure for a particular set of resource constraints is a problem in combinatorial optimization, with a very large search space. Evaluating the results of our optimizations will often require an exhaustive search over all possible representations, which is very time and space expensive. To assist us in this research, we are proposing the purchase of a high-performance computer that will allow us to evaluate our optimization methods as efficiently as possible.