The differentiability of real functions on normed linear space using generalized subgradients (Q1102500): Difference between revisions

Let \(\phi\) be a real-valued function defined on a real normed space X. The authors use the generalized subdifferential \(\partial^{\square}\phi (x)\) introduced by \textit{P. Michel} and \textit{J.-P. Penot} [C. R. Acad. Sc. Paris, Ser. I 298, 269-272 (1984; Zbl 0567.49008)] as a tool in determining differentiability properties of \(\phi\). First, they show that the Gâteaux differentiability of \(\phi\) can be deduced from the Gâteaux differentiability of the norm if \(\phi\) has a directional derivative which attains a constant related to its Michel-Penot generalized directional derivative. Next, it is proved that, if the norm in X is uniformly Gâteaux differentiable, then, for any distance function (generated by a non-empty closed set K in X), the Clarke and Michel-Penot generalized subdifferentials at points off K reduce to the same object. This generates a continuity characterization for Gâteaux differentiability. Moreover, on a Banach space with rotund dual, the Fréchet differentiability of a distance function implies that it is a convex function. Furhter, a mean value theorem for the Michel-Penot subdifferential is obtained and then used to prove some sufficient condition for the directional differentiability of \(\phi\). Another consequence of this mean value theorem is the fact that the monotonicity of the mapping \(x\mapsto \partial^{\square}\phi (x)\) guarantees the convexity of the function \(\phi\).