1: \begin{abstract}

2:   Let $X\in \R^p$ and $Y\in \R$ two random variables. %

3:   In this paper we are interested in the estimation of the conditional

4:   covariance matrix \linebreak $\C\left(\E\left[\boldsymbol{X}\vert

5:       Y\right]\right)$. %

6:   To this end, we will use a plug-in kernel based algorithm. %

7:   Then, we investigate the related performance under smoothness

8:   assumptions on the density function of $(\boldsymbol{X},Y)$. %

9:   Moreover, in high-dimensional context, we shall improve our

10:   estimator to avoid inconsistency issues. %

11:   In the matrix case, the convergence depends on some structural

12:   conditions over the $\C\left(\E\left[\boldsymbol{X}\vert

13:       Y\right]\right)$. %

14: \end{abstract}

15: