1: \begin{abstract} \noindent

2: The Douglas--Rachford splitting algorithm is a 

3: classical optimization method that has found many applications. 

4: When specialized to two normal cone operators, it yields

5: an algorithm for finding a point in the intersection of

6: two convex sets. 

7: This method for solving feasibility problems 

8: has attracted a lot of attention due to its 

9: good performance even in nonconvex settings. 

10: 

11: In this paper, 

12: we consider the Douglas--Rachford algorithm

13: for finding a point in the intersection of two subspaces.

14: We prove that the method converges strongly to the projection of the

15: starting point onto the intersection. Moreover, 

16: if the sum of the two subspaces is closed,

17: then the convergence is linear with the rate being the cosine 

18: of the Friedrichs angle between the subspaces. 

19: Our results improve upon existing results in three ways:

20: First, we identify the location of the limit and thus reveal

21: the method as a best approximation algorithm;

22: second, we quantify the rate of convergence, and third,

23: we carry out our analysis in general (possibly

24: infinite-dimensional) Hilbert space. 

25: We also provide various examples as well as a 

26: comparison with the classical method of alternating projections.

27: \end{abstract}

28: