糖心TV

Dr Ali Gooya, Sheffield University

Abstract: Bayesian probabilistic models provide a systematic approach to combine data with an expert prior knowledge when the estimation of some hidden variables from the observation is required. The incorporation of the prior knowledge in the solution often prevents over and underfitting issues that are commonly faced by non-Bayesian methods. Furthermore, by capturing the causal relationships between different variables, these models provide an explanation for data generation. In this talk, I will present various tasks that we have recently accomplished in our research group, highlighting the Bayesian principles behind each. I will include examples of estimating a probabilistic density function for shapes from pointsets, segmentation of cells in microscopic images using graphical models, and diagnosis of breast cancers in ultrasound images using sparse models. The talk will conclude with some pointers for the future research orientations in this area.