BAYESIAN HYPER‐LASSOS WITH NON‐CONVEX PENALIZATION