Zach Deane-Mayer

That's probably the right place to start! I was thinking of using the `scale` function, and storing the resulting `"scaled:center"` and `"scaled:scale"` attributes, e.g.: ``` r x

That still leaves us with the "un-center" and "un-scale" problem. Another idea would be to use the `caret::preProcess` function, which has nice `predict` methods for new data.

Oh man, it just occurred to me that supporting a `pca` transformation might be super useful too, as it will yield completely un-correlated predictors. It might mess up the print...

Currently [jknowles](https://github.com/jknowles) is leading the charge on this. If you would like to add code, please feel free to submit pull requests on the dev branch.

I have not. That's the reason I skipped multiclass models for now. It seemed non-trival to write a greedy selection function for them. I think you should still be able...

This is also post 1.0. Once we have a working, stable release on CRAN, we'll figure out the multi-class RMSE and AUC optimization functions.

We also need to add multiclass support to caretList and caretStack (caretStack might be easier to do), etc.

The quickest way would be to use `caretStack` with `glmnet` as the stacker. In general, I find myself using `caretStack` more than `caretList`. It's a lot more flexible, and if...

@ajing The real key is the optimization function. I haven't even thought of a hack that will work here.... Once someone's written the multiclass optimization function work, we can work...

We don't support that yet, but it's definitely on the to-do list. For now, you can fit the model separately with a call to `caret::train`, and then add it to...