We are very happy to announce our 11th online seminar in the Biostatistics Seminar Series on Thursday, June 12th, 16h-17h (CET)

This series of Biostatistics seminars targets a broad (bio)statistical audience, in particular PhD-students. Specialists discuss a topic of their interest, paying particular attention to concepts relevant and accessible to a non-specialist audience as well.

Speaker
Wessel van Wieringen
Associate professor of statistics
Department Epidemiology and Data Science, Amsterdam UMC
Department of Mathematics, VU

Title
Regularization: essentials, connections, and applications

Abstract
We discuss regularized estimation in its many disguises. We start by giving a straightforward practical motivation for regularization within the context of the linear regression model. In that context, we introduce the so-called ridge regularization. It is perhaps the most straightforward form of regularization. In our review of ridge regularization, we touch among others upon the bias-variance trade-off, constrained estimation, and Bayesian statistics. Topics that return in the discussion of all other regularization types. We extend ridge regularization framework in two ways. First, we introduce alternative regularization types and point out what they may bring. Secondly, we regularize the estimation of more complicated statistical models, e.g. generalized linear models, nonparametric regression and graphical models, and we mention the difficulties encountered when doing so. Throughout our expose of regularization, we point out connections to regulatory techniques from machine learning such as early stopping, data augmentation, adding noise, and dropout. And at times, we illustrate what regularization may bring through re-analyses of publicly available data from cancer and epidemiology.

Link
T.b.a.