1: \begin{abstract}

2:   We develop model-based methods for solving stochastic convex optimization

3:   problems, introducing the approximate-proximal point, or \aProx, family,

4:   which includes stochastic subgradient, proximal point, and bundle

5:   methods. When the modeling approaches we propose are appropriately

6:   accurate, the methods enjoy stronger convergence and robustness guarantees

7:   than classical approaches, even though the model-based methods typically

8:   add little to no computational overhead over stochastic subgradient

9:   methods.  For example, we show that improved models converge with

10:   probability 1 and enjoy optimal asymptotic normality results under weak

11:   assumptions; these methods are also adaptive to a natural class of what we

12:   term easy optimization problems, achieving linear convergence under

13:   appropriate strong growth conditions on the objective. Our substantial

14:   experimental investigation shows the advantages of more accurate modeling

15:   over standard subgradient methods across many smooth and non-smooth

16:   optimization problems.

17: \end{abstract}

18: