Uncertainty
Reasoning

Laboratory

Practical Bayesian network models are reaching the size of thousands of variables. Also, some real world problems are more naturally represented as hybrid models containing both discrete and continuous variables. Building these models and reasoning with them become increasingly more difficult. The goal of this project is to develop efficient approximate and exact inference algorithms for Bayesian networks and influence diagrams.

Existing multiple-fault diagnosis approaches often produce underspecified explanations that do not fully account for given observation or over-specified explanations that contain redundant variables in explaining the observation. This project aims to provide mathematical formulations for the problem of finding the most relevant faults and develop efficient computational solutions.

In comparison to diagnosis, prognosis aims to predict accurately and precisely the remaining useful life of a failing component or subsystem such that maintenance actions would be scheduled better to reduce system downtime and decrease life-cycle costs dramatically. Prognosis is much less understood and far more challenging. This project plans to develop methodologies for predicting fault evolution to the point that may result in a failure by effectively managing the inherent uncertainty.

This project applies Bayesian modeling and reasoning techniques to various biological problems, such as reverse engineering gene/protein interaction networks from perturbed gene expression datasets and integrating evidence from multiple sources in order to find novel biological patterns and compute their statistical significance levels.

This project aims to utilize the newest advances in artificial intelligence research and develop intelligent collaborative learning and educational assessment systems for promoting students¡¯ more effective learning.

Numerous risk factors can affect the values of supply chain business processes. This project is to develop Bayesian methods for supply chain risk analysis, mitigation, and monitoring.