On some differentiability problems

Publication date 21 February 1993

linear continuous operators normed spaces

Abstract approximation theory (approximation in normed linear spaces and other abstract spaces) (41A65) Fréchet and Gateaux differentiability in optimization (49J50)

Review text

[For the entire collection see Zbl 0746.00073.] Let \(X\), \(Y\) be two normed spaces. Let \(\Phi\) be a class of functions mapping the space \(X\) into the space \(Y\). Let \(\mathcal M\) be a family of subsets in \(X\). Let \(f\) be a function mapping an open set \(\Omega\) contained in \(X\) into \(Y\). Let \(g(t)\) be a real positive valued function defined on the interval \((0,+\infty)\) such that (1): \(\lim_{t\to 0} g(t)=0\). We say that the function \(f\) is \((\Phi,{\mathcal M},g(t))\)- differentiable at the point \(x\in\Omega\) if there is \(\phi\in\Phi\) such that for each \(\varepsilon>0\) and each set \(M\) containing \(x\) belonging to the class \(\mathcal M\) there is a \(\delta>0\) such that for all \(h\) such that \(x+h\in\Omega\cap M\) and such that \(\| h\|<\delta\) we have (2) \(\| f(x+h)-f(x)-[\phi(x+h)-\phi(x)]\|<\varepsilon g(\| h\|)\). If \(\Phi\) is the class of all linear continuous operators mapping the space \(X\) into the space \(Y\) we shall speak briefly about \(({\mathcal M},g(t))\)-differentiability. In this case formula (2) can be rewritten in a simpler form, namely (3) \(\| f(x+h)-f(x)-\phi(h)\|<\varepsilon g(\| h\|)\). The \(({\mathcal M},t)\)-differentiable functions will be called \(({\mathcal M})\)-differentiable. In the paper several classes for \(\mathcal M\) are discussed. Some of them are new. The relations between different types of differentiability are investigated. The definition of \(({\mathcal M})\)-differentiable function is similar to the notion of \(\mathcal M\)-differentiable functions introduced by \textit{V. Averbukh} and \textit{O. Smolyanov} [Usp. Mat. Nauk 22, No. 6(138), 201-220 (Russian) (1967; Zbl 0195.426)]. Their definition is the following: A function is called \(\mathcal M\)-differentiable if for each set \(M\in{\mathcal M}\) and each \(\varepsilon>0\) there is \(\delta>0\) such that for any \(h\in tM\), \(0<t<\delta\), we have (4) \(\| f(x+h)-f(x)- \phi(h)\|<\varepsilon\| h\|\). The essential difference is that \(({\mathcal M})\)-differentiability does not request anything about the family of the sets. In the definition of \(\mathcal M\)-differentiability it is essential that the sets \(M\in{\mathcal M}\) do not contain rays. We show that for many classes of bounded sets (e.g. open bounded sets, compact sets, bounded sets in \(n\)-dimensional subspaces) these two definitions coincide, however, the paper contains a counterexample showing that in general they are different.

Recommendations

Cited in

(17)

This page was built for publication: On some differentiability problems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4026940)