Time: Tuesday 14.03, 12.15–14.00

Room: F235, Kronstad

Speaker: Olivier Verdier, HVL

Title: Graphical Models, Exponential Families, and Variational Inference, part 1

Abstract:

Statistics underlies many engineering activities: as soon as there is noise, statistics is necessary. Modern statistics is based on statistical models (and Bayesian inference). “Exponential families” is a specific subset of statistical models which are incredibly powerful and simple at the same time. They are related to many other scientific areas, such as machine learning, statistical physics, error correcting code, and many more. I will present the basic (amazing) properties of exponential families in this talk.

Based on the book: Graphical Models, Exponential Families, and Variational Inference.