Literature on "pure" supervised feature learning?

I'm interested in literature on supervised feature learning, specifically in an environment when labels are always available. I see that there are a lot of methods for semi-supervised feature learning, but I couldn't find anything for purely supervised feature learning.

The only method I am aware of is to use the hidden layer activations from a neural net or svm trained to predict the targets from the data. Are there any other important methods out there that I'm missing?

To me it seems like this is an under-appreciated problem; any nontrivial supervised learning task should involve supervised feature learning as a first step.