1: \begin{abstract}

2: In this paper, we consider multi-stage stochastic optimization problems with convex objectives and conic constraints at each stage.

3: We present a new stochastic first-order method, namely the dynamic stochastic approximation (DSA) algorithm, for solving these

4: types of stochastic optimization problems. We show that DSA can achieve an optimal  ${\cal O}(1/\epsilon^4)$ rate of convergence

5: in terms of the total number of required scenarios when applied to a three-stage stochastic optimization problem. 

6: We further show that this rate of convergence

7: can be improved to ${\cal O}(1/\epsilon^2)$ when the

8: objective function is strongly convex.  We also discuss variants of DSA for solving more general multi-stage

9: stochastic optimization problems with the number of stages $T > 3$. The developed DSA algorithms only need to go through the scenario tree once in order to

10: compute an $\epsilon$-solution of the multi-stage stochastic optimization problem. To the best of our knowledge,

11: this is the first time that stochastic approximation type methods are generalized for

12: multi-stage stochastic optimization with $T \ge 3$.

13: \end{abstract}

14: