When is L1, L2 or e-net preferred when analyzing data?


It can't all be based on cross-validation?