Consider $\mathcal H_k$ to be the RKHS of a reproducing kernel $k$. I am interested in the differentiability properties of $\mathcal H_k$ as a space of functions.
More precisely, is there a link between the differentiability properties of $k$ and that of the functions in $\mathcal H_k$ ?
In the Gaussian process litterature, there are results about mean square differentiability of the sample paths of Gaussian processes associated to the covariance function $k$. For instance if $k$ is defined on the real line, mean square differentiability of the sample paths amounts to continuous differentiability of $k$. Yet this does not inform us on the properties of the functions in the RKHS since sample paths do not belong to the RKHS in general.
Does anyone know of results on differentiability of the functions in RKHS ? I am especially interested in the case of the Matérn kernel, which is supposed to result in RKHSs of more or less smooth functions depending on the parameter values. But what does "smooth" mean exactly ?