1: \begin{abstract} 

2: \noindent

3: We consider $\min\{f(x):g(x) \le 0, ~x\in X\},$

4: where $X$ is a compact convex subset of $\RR^m$, and $f$ and $g$

5: are continuous convex functions defined on an open neighbourhood

6: of $X$.  We work in the setting of derivative-free optimization,

7: assuming that $f$ and $g$ are available through a black-box that

8: provides only function values for a lower-$\mathcal{C}^2$ representation

9: of the functions.  We present a derivative-free

10: optimization variant of the $\eps$-comirror algorithm \cite{BBTGBT2010}.

11: Algorithmic convergence hinges on the ability to accurately approximate

12: subgradients of lower-$\mathcal{C}^2$ functions, which we prove

13: is possible through linear interpolation. We provide convergence

14: analysis that  quantifies the difference between the function values

15: of the iterates and the optimal function value.  We find that the

16: DFO algorithm we develop has the same convergence result as the

17: original gradient-based algorithm.  We present some 

18: numerical testing that demonstrate the practical feasibility of the

19: algorithm, and conclude with some directions for further research.

20: \end{abstract}

21: