1: \begin{abstract} We introduce an inexact variant of Stochastic Mirror Descent (SMD), called Inexact Stochastic Mirror Descent (ISMD), to solve

2: nonlinear two-stage stochastic programs where the second stage problem 

3: has linear and nonlinear coupling constraints and a 

4: nonlinear objective function which depends on both first and second

5: stage decisions. Given a candidate first stage solution and a realization of the second stage random vector, each iteration of ISMD combines a stochastic subgradient descent using a prox-mapping

6: with the computation of approximate (instead of exact for SMD) primal and dual second stage solutions.

7: We provide two convergence analysis of ISMD, under two sets of assumptions.

8: We show in particular that under some assumptions, ISMD has the same convergence rate as SMD. 

9: The first convergence analysis is based on the formulas for inexact cuts of value functions of

10: convex optimization problems shown recently in \cite{guigues2016isddp}.

11: The second convergence analysis relies on

12: new formulas that we derive for inexact cuts of value functions of convex optimization problems 

13: assuming that the dual function of the second stage problem for all fixed first stage 

14: solution and realization of the second stage random vector, is strongly concave. 

15: We show that this assumption of strong concavity is satisfied for some classes of problems and present the

16: results of numerical experiments on two simple two-stage problems which show that 

17: solving approximately the second stage problem for the first iterations of ISMD can help us obtain a

18: good approximate first stage solution quicker than with SMD.\\

19: \end{abstract}

20: