1: \begin{abstract}

2:     We present \nccsg{}, a method for constrained nonsmooth convex optimization.

3:     In each iteration, \nccsg{} finds the best norm-constrained descent

4:     direction by considering the worst bound over all local subgradients. We

5:     prove a few global convergence rates of \nccsg{}. For well-behaved nonsmooth

6:     functions (characterized by the weak smoothness property and being Lipschitz

7:     continuous), \nccsg{} converges in $\bigO{\epsilon^{-1}}$ iterations, where

8:     $\epsilon$ is the desired optimality gap. \nccsg{} converges in

9:     $\bigO{\epsilon^{-0.5}}$ iterations for strongly convex, weakly smooth

10:     functions. Furthermore, if the function is strongly convex and smooth, then

11:     \nccsg{} achieves linear convergence (i.e., $\bigO{-\log \epsilon}$). The

12:     overall efficiency of \nccsg{} depends on the efficiency of solving a

13:     minimax optimization problem involving the subdifferential of the objective

14:     function in each iteration.

15: \end{abstract}

16: