Feature request: support adaptive penalty weights?

[tomwenseleers]

I was wondering if it would be hard to allow NNLM to accept adaptive penalty weights, so one could use it to carry out adaptive LASSO or adaptive ridge (as one can do with the penalty.factor argument in glmnet)? It's just that I was keen to try NNLM as a backend to fit L0-penalized models, which can be done using an iterative adaptive LASSO or iterative adaptive ridge procedure, but that would require adaptive penalty weights...