The classical inverse problem arises in the situation where the system and a set of outputs are known and we are interested in recovering the inputs, and this problem has wide application in system identification, medical imaging, remote sensing and many others. In this thesis, in addition to the classical view, we will extend the inverse problem to the situation where multiple pairs of inputs and outputs are known, and we are interested in recovering the system (functional relationship) between the pair, and this will include statistical problems such as correlation analysis and regression. We will discuss statistical models and inference algorithms (MCMC, variational inference, optimization based) designed for various scientific fields ranging from nuclear non-proliferation to microbial systems to extrapolate meaningful information of interest.
Speaker(s): Haonan Zhu,
Virtual: https://events.vtools.ieee.org/m/367550