New Lab-Length lecture on Generalized Additive Models

[davidrosenberg]

These are fantastic for building interpretable nonlinear models. People love them when they learn about them. Shockingly absent from most courses. One could actually build a version of them with our tools through gradient boosting, by restricting the base hypothesis space to functions that depend only on one feature. Though that would not be the traditional approach.