1: \begin{abstract}

2: We examine an application of the kernel-based interpolation 

3: to numerical solutions for Zakai equations in nonlinear filtering, and aim to prove its 

4: rigorous convergence. To this end, we find the class of kernels and the structure of 

5: collocation points explicitly under which the process of iterative interpolation is stable. 

6: This result together with standard argument in error estimation shows that 

7: the approximation error is bounded by the order of the square root of the time step and 

8: the error that comes from a single step interpolation. 

9: Our theorem is well consistent with the results of numerical experiments. 

10: 

11: \begin{flushleft}

12: {\bf Key words}: 

13: Zakai equations, kernel-based interpolation,  

14: stochastic partial differential equations, radial basis functions. 

15: \end{flushleft}

16: \begin{flushleft}

17: {\bf AMS MSC 2010}: 

18: 60H15, 65M70, 93E11.

19: \end{flushleft}

20: \end{abstract}

21: