At the risk of making the module on linear models too long, I still have the feeling that the "Regularization in linear model" section could have an exercise on regression and/or a notebook on classification. My proposal is this:
- Add a completely new notebook on regression
- Transform the current classification exercise into a notebook
- Add a new exercise on classification to replace the previous one
I think we can showcase how the optimal regularization parameter depends on the scoring metric passed to the grid/random search. This can be the new exercise for either regression or classification.