1: \begin{abstract}

2:   We consider estimation of conditional hazard functions and densities over the

3:   class of multivariate càdlàg functions with uniformly bounded sectional

4:   variation norm when data are either fully observed or subject to

5:   right-censoring. We demonstrate that the empirical risk minimizer is either

6:   not well-defined or not consistent for estimation of conditional hazard

7:   functions and densities. Under a smoothness assumption about the

8:   data-generating distribution, a highly-adaptive lasso estimator based on a

9:   particular data-adaptive sieve achieves the same convergence rate as has been

10:   shown to hold for the empirical risk minimizer in settings where the latter is

11:   well-defined. We use this result to study a highly-adaptive lasso estimator of

12:   a conditional hazard function based on right-censored data. We also propose a

13:   new conditional density estimator and derive its convergence rate. Finally, we

14:   show that the result is of interest also for settings where the empirical risk

15:   minimizer is well-defined, because the highly-adaptive lasso depends on a much

16:   smaller number of basis function than the empirical risk minimizer.

17: \end{abstract}

18: