1: \begin{abstract}

2:   We study acceleration and preconditioning strategies for

3:   a class of Douglas--Rachford methods aiming at the solution

4:   of convex-concave saddle-point problems associated with

5:   Fenchel--Rockafellar duality. While the basic iteration converges

6:   weakly in Hilbert space with $\mO(1/k)$ ergodic convergence of

7:   restricted primal-dual gaps, acceleration can be achieved under

8:   strong-convexity assumptions. Namely, if either the primal or dual

9:   functional in the saddle-point formulation

10:   is strongly convex, then the method can be modified to

11:   yield $\mO(1/k^2)$   ergodic convergence. 

12:   In case of both functionals

13:   being strongly convex, similar modifications lead to an asymptotic

14:   convergence of $\mO(\vartheta^k)$ for some $0 < \vartheta < 1$.

15:   All methods allow in particular for preconditioning, i.e., the

16:   inexact solution of the implicit linear step in terms of linear

17:   splitting methods with all convergence rates being maintained.

18:   The efficiency of the proposed methods is verified and compared

19:   numerically, especially showing competitiveness with

20:   respect to state-of-the-art accelerated algorithms.

21: \end{abstract}

22: