1: \begin{abstract}

2: We study multi-parameter Tikhonov regularization, i.e., with multiple penalties. Such

3: models are useful when the sought-for solution exhibits several distinct features

4: simultaneously. Two choice rules, i.e., discrepancy principle and balancing principle,

5: are studied for choosing an appropriate (vector-valued) regularization parameter, and

6: some theoretical results are presented. In particular, the consistency of the discrepancy

7: principle as well as convergence rate are established, and an a posteriori error estimate

8: for the balancing principle is established. Also two fixed point algorithms are proposed

9: for computing the regularization parameter by the latter rule. Numerical results for

10: several nonsmooth multi-parameter models are presented, which show clearly their superior

11: performance over their single-parameter counterparts.

12: \\

13: Key words: multi-parameter regularization, value function, balancing principle, parameter

14: choice.

15: \end{abstract}

16: