Model Complexity

I think it's a very interesting paper, but it seems to only use Ridge to get a result. I was expecting they would have used some deep net, and then instead of doing Ridge type shrinkage, just adjust weight decay upwards.