1: \begin{abstract}

2: Regularized kernel methods such as support vector machines (SVM)

3: and support vector regression (SVR)

4: constitute a broad and flexible class of

5: methods which are theoretically well investigated and 

6: commonly used in nonparametric classification and regression problems.

7: As these methods are based on a Tikhonov regularization

8: which is also common in inverse problems,

9: this article investigates the use of 

10: regularized kernel methods for inverse problems in a unifying way.

11: Regularized kernel methods are based on the use of reproducing kernel Hilbert spaces

12: (RKHS) which lead to very good computational properties.

13: It is shown that similar properties remain true in solving 

14: statistical inverse problems and that standard software implementations

15: developed for ordinary regression problems can still be used for

16: inverse regression problems. 

17: 

18: Consistency of these methods and

19: a rate of convergence for the risk is shown under quite 

20: weak assumptions and

21: rates of convergence for the estimator are 

22: shown under somehow stronger assumptions.

23: The applicability of these methods is demonstrated in a simulation.

24: \end{abstract}