Variational Bayes: An integral part of (modern) statistics

Prof. Haavard Rue (King Abdullah University of Science and Technology)

Friday 17th January 11:00-12:00 Maths and Stats:311B

Abstract

In this talk, I will discuss Variational Bayes (VB), what it is, why we should
care, and why it should be an integral part of any (modern) statistician's
toolbox. I will discuss our use of it in the R-INLA project, but also its role in the 'The Bayesian Learning Rule' from which a wide-range of algorithms can be derived from: ridge regression, Newton's method, Kalman filter, stochastic-gradient descent, RMSprop, and Dropout, Laplace's method, EM and so on.

Add to your calendar

Download event information as iCalendar file (only this event)