Understanding Bayesian Boosting: Convincing Models Beyond Linear Assumptions
I recently watched a presentation by George Perrett on Bayesian Boosting, which clarified some essential concepts in modeling uncertainty through non-linear approaches. Perrett starts by discussing his high school experience with AP Statistics, where he grappled with methods that assumed linear relationships. Fast forward a decade, and he now advocates for methods that embrace non-linearity, specifically Bayesian models. The Problem of Assumptions He opens with an example that brings out a key distinction between expected values and actual distributions.