Integrability of subdifferentials of certain bivariate functions. (Q1406776): Difference between revisions

Integrability of subdifferentials refers to the question under what conditions the inclusion \(\partial\varphi(x)\subset \partial\psi(x)\) implies that the functions \(\varphi\) and \(\psi\) differ (locally) by a constant only. In the present paper, the subdifferential mapping \(x\mapsto \partial\varphi(x)\) is defined axiomatically such that for \(\varphi\) convex it coincides with the subdifferential of convex analysis and that, roughly speaking, it is able to detect local minimizers of \(\varphi\). In particular, the generalized Clarke subdifferential of a not necessarily locally Lipschitz function [see \textit{R. T. Rockafellar}, Can. J. Math. 32, 257--280 (1980; Zbl 0447.49009)] is a subdifferential in the above sense. The authors consider bivariate functions of the form \(f,g: X\times Y\to \mathbb{R}\cup\{+\infty\}\), where \(X\) and \(Y\) denote Banach spaces. Introducing certain regularity and directional Lipschitz properties, they establish integrability results both in terms of partial subdifferentials and in terms of ``joint'' subdifferentials.