1: \begin{abstract}

2: Stochastic approximation techniques have been used in various

3: contexts in data science. We propose a stochastic version of the

4: forward-backward algorithm for minimizing the sum of two

5: convex functions, one of which is not necessarily smooth. Our

6: framework can handle stochastic approximations of the

7: gradient of the smooth function and allows for stochastic errors in

8: the evaluation of the proximity operator of the nonsmooth function.

9: The almost sure convergence of the iterates generated

10: by the algorithm to a minimizer 

11: is established under relatively mild assumptions. We also

12: propose a stochastic version of a popular primal-dual proximal

13: splitting algorithm, establish its convergence, and apply it to

14: an online image restoration problem.

15: \end{abstract}

16: