1: \begin{abstract}

2: Error estimates for kernel interpolation in Reproducing Kernel Hilbert Spaces (RKHS) usually assume quite restrictive properties on the shape of the domain, 

3: especially in the case of infinitely smooth kernels like the popular Gaussian kernel. 

4: 

5: In this paper we leverage an analysis of greedy kernel algorithms to prove that it is possible to obtain convergence results (in the number of interpolation 

6: points) for kernel interpolation for arbitrary domains $\Omega \subset \R^d$, thus allowing for non-Lipschitz domains including e.g.\ cusps and irregular boundaries. %

7: Especially we 

8: show that, when going to a smaller domain $\Omegat \subset \Omega \subset \R^d$, the convergence rate does not deteriorate --- i.e.\ the convergence rates are stable with respect to going to a subset.

9: 

10: The impact of this result is explained on the examples of kernels of finite as well as infinite smoothness like the Gaussian kernel. A comparison to 

11: approximation in Sobolev spaces is drawn, where the shape of the domain $\Omega$ has an impact on the approximation properties.

12: Numerical experiments illustrate and confirm the experiments. %

13: %

14: \end{abstract}

15: