1: \begin{abstract}

2: We consider monotone inclusion problems where the operators may be

3: expectation-valued, a class of problems that subsumes convex stochastic

4: optimization problems as well as subclasses of stochastic variational

5: inequality and equilibrium problems. A direct application of splitting

6: schemes is complicated by the need to resolve problems with

7: expectation-valued maps at each step, a concern that is addressed by

8: using sampling. Accordingly, we propose an avenue for addressing

9: uncertainty in the mapping: {\em Variance-reduced stochastic modified

10: forward-backward splitting scheme} ({\bf vr-SMFBS}). In constrained

11: settings, we consider structured settings when the map can be

12: decomposed into an expectation-valued map $A$ and a maximal monotone

13: map $B$ with a tractable resolvent.  We show that the proposed schemes

14: are equipped with a.s. convergence guarantees, linear (strongly

15: monotone $A$) and $\mathcal{O}(1/k)$ (monotone $A$) rates of

16: convergence while achieving optimal oracle complexity bounds. 

17: The rate statements in monotone regimes appear to be amongst the first

18: and rely on leveraging the Fitzpatrick gap function for monotone

19: inclusions. Furthermore, the schemes rely on weaker moment requirements

20: on noise and allow for weakening unbiasedness requirements on

21: oracles in strongly monotone regimes. Preliminary numerics on a class

22: of two-stage stochastic variational inequality problems reflect these

23: findings and show that the variance-reduced schemes outperform

24: stochastic approximation schemes and sample-average approximation

25: approaches. The benefits of attaining deterministic

26: rates of convergence become even more salient when resolvent

27: computation is expensive.

28: \end{abstract}

29: