The paper "Predictive and interpretable models via the stacked elastic net" stacks multiple elastic nets to train a well-interpretable meta-learner that often outperforms an elastic net with tuned penalty parameters. It has been implemented as the R-package starnet.