1: \begin{abstract}

2: We give a comprehensive theoretical characterization of a nonparametric estimator for the $L_2^2$ divergence between two continuous distributions. 

3: We first bound the rate of convergence of our estimator, showing that it is $\sqrt{n}$-consistent provided the densities are sufficiently smooth.

4: In this smooth regime, we then show that our estimator is asymptotically normal, construct asymptotic confidence intervals, and establish a Berry-Ess\'{e}en style inequality characterizing the rate of convergence to normality.

5: We also show that this estimator is minimax optimal.

6: 

7: %% We establish a number of theoretical properties for a nonparametric estimator of the $L_2^2$-divergence between two continuous distributions.

8: %% show that our estimator is $\sqrt{n}$-consistent when the smoothness of the densities $s$ is at least $d/4$ where $d$ is the dimension.

9: %% We present some applications demonstrating the empirical appeal of our estimator. 

10: %% \ifthenelse{\equal{\version}{image}}

11: %% {We present an application to object recognition problems in computer vision, showing that using learning algorithms with a similarity kernel based on our estimator is competitive with state-of-the-art methods.}

12: %% {We present a number of numerical simulations confirming our theoretical results and comparing our estimator to other methods for two sample testing. %%and distribution-based machine learning.

13: %% We show our estimator to other methods for two sample testing and machine learning over distributions

14: %% }

15: \end{abstract}

16: