Seminar Abstract

“Advanced Probabilistic Network Modeling with Qualitative Prior Knowledge and Application in Quantitative Inferring Multi-scale Molecular Interactions Network”

In this talk, I will propose unprecedented solutions to the challenges in Bayesian network learning, namely, how to construct prior distribution over structure and parameter space from prevalent amount of preexisting qualitative information in science and industrial domain within an unified framework as well as to the tough question how qualitative statements about relationship between domain entities can be transformed to yield quantitative predictive models, able to perform probabilistic inference and reasoning. To this end, we will only consider the statistics and uncertainty presented by prior information, i.e. we utilize solely qualitative prior information in our study and therefore, no quantitative data information is available to shield our insights in the function and effects of prior knowledge in probabilistic modeling with Bayesian networks.