1: \begin{abstract}

2: A hypodifferential is a compact family of affine mappings that defines a local max-type approximation of a nonsmooth

3: convex function. We present a general theory of hypodifferentials of nonsmooth convex functions defined on a Banach

4: space. In particular, we provide a complete characterization of hypodifferentiability and hypodifferentials of nonsmooth

5: convex functions, derive calculus rules for hypodifferentials, and study the Lipschitz continuity/Lipschitz

6: approximation property of hypodifferentials that can be viewed as a natural extension of the Lipschitz continuity of 

7: the gradient to the general nonsmooth setting. As an application of our theoretical results, we study the rate of

8: convergence of several versions of the method of hypodifferential descent for nonsmooth convex optimization and present

9: an accelerated version of this method having the faster rater of convergence $\mathcal{O}(1/k^2)$.

10: \end{abstract}

11: