Campus, PAB 232
Join us Friday Feb 14th at 3pm in PAB 232 on campus (and on zoom ) for the next meeting of the Stats and ML Journal Club. This week, I (Sean McLaughlin) will be leading an introduction on Gaussian Processes. See you all then!
Title: Everything you wanted to know (and many things you didn't) about Gaussian Processes
Abstract: Gaussian processes are the gold standard of non-parametric regression models. By conditioning the space of all functions on our data, one can simultaneously obtain a good fit, uncertainties, and samples of all possible functions that fit the data. However, their numerical overheads can make their practical use difficult. Still, there are plenty of tricks that make them applicable to real science problems. Additionally, many cutting edge Bayesian learning methods seek to approximate the behavior of GPs, so decent theoretical understanding is even more relevant than ever. I will overview their basic theoretical concepts, practical regression methods, and related ML models of Gaussian Processes.