1: \begin{abstract}

2: We revisit the classical Douglas-Rachford (DR) method for finding a zero of the  sum of two maximal monotone operators.  

3: Since the practical performance of  the DR method  crucially depends on the stepsizes, we aim at developing an adaptive stepsize rule.  

4: To that end, we  take a closer look at a linear case of the problem and use our  findings to develop a stepsize strategy that eliminates the need for stepsize tuning.  

5: We analyze a general non-stationary DR scheme  and prove its convergence for a convergent sequence of stepsizes with summable increments in the case of maximally monotone operators.

6: This, in turn, proves the convergence of the method with the new adaptive stepsize rule.  

7: We also derive the related non-stationary alternating direction method of multipliers (ADMM)

8: We illustrate the efficiency of the proposed methods on several numerical examples.

9: 

10: \end{abstract}

11: