On uniformly convex functions

From MaRDI portal



Abstract: Non-convex functions that yet satisfy a condition of uniform convexity for non-close points can arise in discrete constructions. We prove that this sort of discrete uniform convexity is inherited by the convex envelope, which is the key to obtain other remarkable properties such as the coercivity. Our techniques allow to retrieve Enflo's uniformly convex renorming of super-reflexive Banach spaces as the regularization of a raw function built from trees. Among other applications, we provide a sharp estimation of the distance of a given function to the set of differences of Lipschitz convex functions. Finally, we prove the equivalence of several natural fashions to quantify the non-super weakly compactness of a subset of a Banach space.


Uniform convexity of the norm of a Banach space is a classical geometrical notion introduced by \textit{J. A. Clarkson} [Trans. Am. Math. Soc. 40, 396--414 (1936; Zbl 0015.35604)]. It implies reflexivity. A cornerstone renorming result related to this notion is due to \textit{P. Enflo} [Isr. J. Math. 13, 281--288 (1973; Zbl 0259.46012)]: A Banach space is superreflexive if, and only if, it has an equivalent uniformly convex norm. Uniform convexity for extended-valued functions was considered by \textit{E. S. Levitin} and \textit{B. T. Polyak} [Sov. Math., Dokl. 7, 764--767 (1966; Zbl 0161.07002); translation from Dokl. Akad. Nauk SSSR 168, 997--1000 (1966)]. A central concept in the present paper is a relaxed condition for (non-necessarily convex) functions, namely \(\varepsilon\)-uniform convexity with respect to a pseudometric \(d\) for some \(\varepsilon>0\) (i.e., there exists \(\delta>0\) such that, whenever \(d(x,y)\ge\varepsilon\), then \(f(\frac{x+y}{2})\le \frac{f(x)+f(y)}{2} -\delta\)). The paper discusses, first, the basic properties of \(\varepsilon\)-uniformly convex and \(\varepsilon\)-uniformly quasiconvex (replace the second term in the previous inequality by \(\max\{f(x),f(y)\}-\delta\)) functions; second, it shows a key reduction-to-convexity argument (Theorem~1.3) that allows for getting back to the convex case; third, a section is devoted to constructing uniformly convex functions from the non-existence of some trees. General properties of \(\varepsilon\)-uniformly convex functions are treated later (even adding conditions like Lipschitzness or homogeneity). It continues by proving an estimation of the approximation by differences of convex functions. Several measures of super weak noncompactness (a notion introduced in [\textit{M.~Raja}, J. Math. Anal. Appl. 439, No.~1, 183--196 (2016; Zbl 1351.46010)]) are proposed. The paper ends by providing a proof of Enflo's theorem quoted above based on the new ideas introduced. The techniques used here are of interest beyond the last application. It is worth to mention Theorem~1.4: Let \(f\) be a convex proper lowersemicontinuous \(\varepsilon\)-uniformly convex extended-valued function on a Banach space \((X,{\|\cdot\|})\). Then there exists an equivalent norm \({|||\cdot |||}\) on \(X\) such that the function \(x\mapsto ||| x|||^2\) is \(\varepsilon'\)-uniformly convex for every \(\varepsilon'>\varepsilon\) on the subsets of \(\mathrm{dom}(f)\) where \(f\) is bounded above. Moreover, \({|||\cdot |||}\) can be taken as close to \({\|\cdot\|}\) as we wish.



Cites work


Cited in
(22)








This page was built for publication: On uniformly convex functions

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5919121)