Describir: Selecting Penalty Parameters of High-Dimensional M-Estimators Using Bootstrapping after Cross Validation.