Presentation on Bayesian Learning by Pradeep Gowda First of a multi-part series on Bayesian analysis Covered some basic background and then focused in on a analysis of conditional inferences of positive cancer screenings. Presentation on Generalization by Sheamus Parkes Very Basic Notes: Classic bias variance curve Best algorithms let you walk the tradeoff It's all just ridging Random forest is a bit different Subtle overfitting sources… Tuning one (or two) hyperparameters won't overfit... Tuning many algorithms and picking the best will overfit Not including feature selection as part of the tuning process Links Mentioned: Credibility, penalized regression and boosting: let's call the whole thing off: https://www.casact.org/education/infocus/2011/handouts/AM2-Fry.pdf glmnet package for R http://cran.r-project.org/web/packages/glmnet/index.html http://www.jstatsoft.org/v33/i01/