1: \begin{abstract}

2:   The lasso procedure is ubiquitous in the statistical and signal

3:   processing literature, and as such, is the target

4:   of substantial theoretical and applied

5:   research. While much of this research focuses on the desirable

6:   properties that lasso possesses---predictive risk consistency,

7:   sign consistency, correct model selection---all of it has assumes

8:   that the tuning parameter is chosen in an oracle fashion. Yet, this

9:   is impossible in practice. Instead, data analysts must use the

10:   data twice, once to choose the tuning parameter and again to

11:   estimate the model. But only heuristics have ever justified such

12:   a procedure. To this end, we give the first definitive answer

13:   about the risk consistency of lasso when the smoothing parameter is

14:   chosen via cross-validation. We show that under some restrictions on the

15:   design matrix, the lasso estimator is still risk consistent with an

16:   empirically chosen tuning parameter. \\

17:   \vspace{1em}

18: 

19:   \noindent{\bf Keywords:} stochastic equicontinuity, uniform convergence,

20:     persistence

21: \end{abstract}

22: