1: \begin{abstract}

2: The celebrated Nadaraya-Watson kernel estimator is among the most studied method for nonparametric regression. A classical result is that its rate of convergence depends on the number of covariates and deteriorates quickly as the dimension grows, which underscores the ``curse of dimensionality'' and has limited its use in high dimensional settings. In this article,  we show that 

3: when the true regression function is single

4: or multi-index, the effects of the curse of dimensionality may be mitigated for the Nadaraya-Watson kernel estimator.  Specifically, we prove that with $K$-fold cross-validation,

5: the Nadaraya-Watson kernel estimator indexed by a positive semidefinite bandwidth matrix

6: has an oracle property that its rate of convergence depends

7: on the number of indices of the regression function rather than the number of covariates.  

8: Intuitively, this oracle property is a consequence of allowing the bandwidths to diverge to infinity as opposed to restricting them all to converge to zero at certain rates as done in previous theoretical studies.

9: Our result provides a theoretical perspective for the use of kernel estimation in high dimensional nonparametric regression and other applications such as metric learning when a low rank structure is anticipated.

10: Numerical illustrations are given through simulations and real data examples. 

11: \end{abstract}

12: