1: \begin{abstract}

2: We address the problem of approximating an unknown function from its

3: discrete samples given at arbitrarily scattered sites. This problem is

4: essential in numerical sciences, where modern applications also

5: highlight the need for a solution to the case of functions with

6: manifold values. In this paper, we introduce and analyze a combination

7: of kernel-based quasi-interpolation and multiscale approximations for

8: both scalar- and manifold-valued functions. While quasi-interpolation

9: provides a powerful tool for approximation problems if the data is

10: defined on infinite grids, the situation is more complicated when it

11: comes to scattered data. \hw{Here, higher-order quasi-interpolation

12:   schemes either require derivative information or become numerically

13:   unstable.} Hence, this paper \rev{principally studies} the

14: improvement achieved by combining quasi-interpolation with a

15: multiscale technique. The main contributions of this paper are as

16: follows. First, we introduce the multiscale quasi-interpolation

17: technique for scalar-valued functions. Second, we show how this

18: technique can be carried over \rev{using moving least-squares

19:   operators} to the manifold-valued setting. Third, we give a

20: mathematical proof that converging quasi-interpolation will also lead

21: to converging multiscale quasi-interpolation. Fourth, we provide ample

22: numerical evidence that multiscale quasi-interpolation has superior

23: convergence to quasi-interpolation. In addition, we will provide

24: examples showing that the multiscale quasi-interpolation approach

25: offers a powerful tool for many data analysis tasks, such as denoising

26: and anomaly detection. It is especially attractive for cases of

27: massive data points and high dimensionality.   

28: \end{abstract}

29: