Max Kuhn

Hmm. My chapter on categorical predictors encodes these as unknown. Maybe they should be actually missing values so that effect encodings work without estimating their level. _Originally posted by @topepo...

There is currently a button to run the computations manually. The computations probably aren't that slow so remove that button and use basic shiny (unless it is too slow in...

clean up/simply some of the Ames and hotel rate code

I will eventually write summaries of other methods (mostly for deep learning) like hyperopt, spearmint, etc. Not a high priority right now. _Originally posted by @topepo in https://github.com/aml4td/website/pull/57#discussion_r1763580121_

See tidymodels/parsnip#1187

`contr_one_hot()` is also in parsnip (where it is exported). There are tests there that are a bit more extensive: https://github.com/tidymodels/parsnip/blob/main/tests/testthat/test-contr_one_hot.R You can bring those over here and I'll open a...

update predict.modelfit(type = "quantile")

6

We are adding a mode for quantile regression but have one engine that already enables such prediction (using the censored regression mode). We should allow that but make some adjustments...

It requires an `augment_quantile()` function ([examples](https://github.com/tidymodels/parsnip/blob/main/R/augment.R#L88))

Brulee neural networks with two hidden layers

1

A complement to tidymodels/brulee#84 and the previous tidymodels/brulee#80 Tests will be in extratests