Metric entropy of classes of sets with positive reach (Q1745370)

The notion of metric entropy (also called \(\varepsilon\)-entropy) has been invented by \textit{A. N. Kolmogorov} [Dokl. Akad. Nauk SSSR 108, 385--388 (1956; Zbl 0070.11501)], in order to classify compact metric sets according to their massivity. Questions in the theory of metric entropy and approximation were studied by \textit{A. N. Kolmogorov} and \textit{V. M. Tikhomirov} [Transl., Ser. 2, Am. Math. Soc. 17, 227--364 (1959; Zbl 0133.06703)], \textit{G. F. Clements} [Pac. J. Math. 13, 1085--1095 (1963; Zbl 0158.05002)], and by \textit{G. G. Lorentz} [Bull. Am. Math. Soc. 72, 903--937 (1966; Zbl 0158.13603)]. \textit{R. M. Dudley} extended their results in [J. Approx. Theory 10, 227--236 (1974; Zbl 0275.41011)]. Two years later, \textit{I. N. Bronstein} [``\(\varepsilon\)-entropy of convex sets and functions'', Sibirsk. Mat. Z. 17, No. 3, 508--514 (1976)] obtained metric entropy estimates under Hausdorff distance of classes of bounded convex sets. However, in applications, for instance in image reconstruction of brain activities, it is not realistic to assume that an active region is convex, thus Bronstein's theorem can not be applied. In this article, Bronstein's result is extended to another class of sets, called sets with positive reach. They were introduced by \textit{H. Federer} [Trans. Am. Math. Soc. 93, 418--491 (1959; Zbl 0089.38402)]. The greater generality of these sets make them useful in applications. In this article, the authors prove that, on the one hand, adding a small ball to a pair of sets with positive reach will not change their Hausdorff distance and, on the other hand, adding a small ball to a set with positive reach will smooth its boundary. Based on that, the authors construct a polytope to approximate a fixed set with positive reach and they estimate covering numbers locally. Finally, they iterate the local estimates to obtain sharp metric entropy estimates under Hausdorf distance for classes of bounded sets with positive reach.