Widths and entropy numbers of embeddings of Sobolev classes on a Hölder domain The paper was published with the financial support of the Moscow Center for Fundamental and Applied Mathematics at Lomonosov Moscow State University under the agreement 075-15-2025-345.
A.A. Vasil’eva · 2025
Read Widths and entropy numbers of embeddings of Sobolev classes on a Hölder domain The paper was published with the financial support of the Moscow Center for Fundamental and Applied Mathematics at Lomonosov Moscow State University under the agreement 075-15-2025-345. by A.A. Vasil’eva (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2512-00620.txt.
Introduction
A Vasil'eva wrote a 2025 pure math paper on Sobolev classes, which are sets of functions sorted by how smooth they are. The setting is a domain whose border follows a Hölder condition, a rule that limits how rough the edge can be. The paper measures widths and entropy numbers for ways these classes sit inside other spaces.
Entropy numbers say, roughly, how well a set can be covered by small balls, and widths say how well it can be approximated. Vasil'eva improves older upper estimates for the entropy numbers. Under some conditions on the parameters, the new rate of decrease is sharper than before.
Catalog record
| CREATOR | A.A. Vasil’eva |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 92 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2512-00620
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Widths and entropy numbers of embeddings of Sobolev classes on a Hölder domain The paper was published with the financial support of the Moscow Center for Fundamental and Applied Mathematics at Lomonosov Moscow State University under the agreement 075-15-2025-345. A.A. Vasil’eva Abstract 1 Introduction In [ 1 ] upper estimates for the entropy numbers of Sobolev classes on a domain with Hölder boundary were obtained. In the present paper we improve these estimates (in the case when the definition of the Sobolev class involves all partial derivatives of order r ): we will show that the exponential in the orders of decrease is greater than the quantity in [ 1 ] ; moreover, under some conditions on the parameters, the order will be the same as for domains with Lipschitz boundary (while the boundary of the domain may be non-Lipschitz). We also obtain upper estimates for the Kolmogorov, linear and the Gelfand widths. In addition, we will consider domains of special form, where the Hölder singularity is concentrated on an h -set (see the definition below). For them, we will obtain more sharp upper estimates of widths and entropy numbers. The same lower estimates will be obtained in the case of special h -sets constructed in [ 2 ] . The last example shows that the exponent in the upper estimates on the class of all Hölder domains cannot be improved. On the other hand, it will be shown that these lower estimates hold not for all h -sets; for example, the order of decreae of entropy numbers and widths can be improved if a plane is parallel to a coordinate plane. Let us give necessary definitions. Let d\in\mathbb{N} , and let \Omega\subset\mathbb{R}^{d} be a bounded domain. Given f\in L_{1}^{{\rm loc}}(\Omega) and \alpha=(\alpha_{1},\,\dots,\,\alpha_{d})\in\mathbb{Z}_{+}^{d} , we denote |\alpha|=\alpha_{1}+\dots+\alpha_{d} , \partial_{\alpha}f=\frac{\partial^{|\alpha|}f}{\partial x_{1}^{\alpha_{1}}\dots\partial x_{d}^{\alpha_{d}}} (the partial derivative is taken in the sense of distributions). Given r\in\mathbb{Z}_{+} , we denote by \nabla^{r}f the vector of all partial derivatives \partial_{\alpha}f such that |\alpha|=r , and set \|\nabla^{r}f\|_{L_{p}(\Omega)}=\max_{|\alpha|=r}\|\partial_{\alpha}f\|_{L_{p}(\Omega)}. Let 1\leqslant p\leqslant\infty , r\in\mathbb{N} , and let \Omega\subset\mathbb{R}^{d} be a bounded domain. We define the Sobolev spaces by the formulas {\cal W}^{r}_{p}(\Omega)=\{f\in L_{1}^{{\rm loc}}(\Omega):\;\|\nabla^{r}f\|_{L_{p}(\Omega)}<\infty\}, \tilde{\cal W}^{r}_{p}(\Omega)=\{f\in L_{1}^{{\rm loc}}(\Omega):\;\|\nabla^{r}f\|_{L_{p}(\Omega)}+\|f\|_{L_{p}(\Omega)}<\infty\} and the unit balls (the Sobolev classes), by the formulas W^{r}_{p}(\Omega)=\{f\in L_{1}^{{\rm loc}}(\Omega):\;\|\nabla^{r}f\|_{L_{p}(\Omega)}\leqslant 1\}, \tilde{W}^{r}_{p}(\Omega)=\{f\in L_{1}^{{\rm loc}}(\Omega):\;\|\nabla^{r}f\|_{L_{p}(\Omega)}+\|f\|_{L_{p}(\Omega)}\leqslant 1\}. It is well-known (see, e.g., [ 3 ] ) that if the domain \Omega is a finite unuion of bounded domains with Lipschitz boundary, then for r+\frac{d}{q}-\frac{d}{p}>0 , for r+\frac{d}{q}-\frac{d}{p}=0 , q<\infty , or for r=d , p=1 , q=\infty , the set \tilde{W}^{r}_{p}(\Omega) is bounded in L_{q}(\Omega) (and W^{r}_{p}(\Omega) is the sum of a bounded set and the space of polynomials of degree at most d-1 ). The same assertion holds if \Omega is a John domain [ 4 , 5 , 6 ] or satisfies a decaying flexible cone condition [ 7 ] . Embedding theorems for the Sobolev spaces on various irregular domains with zero angles (e.g., on Hölder domains, on domains with a flexible \sigma -cone condition an so on) were obtained in [ 8 , 9 , 10 , 11 , 12 , 13 , 14 , 15 , 16 , 17 , 18 , 19 , 20 ] . Now we give the definitions of the Gelfand, Kolmogorov and linear widths. Let X be a linear normed space. Denote by X^{*} the dual space to X , by {\cal L}_{n}(X) , the family of all linear subspaces in X of dimension at most n , and by L(X,\,Y) , the space of linear continuous operators A:X\to Y . Given A\in L(X,\,X) , we denote by {\rm rk}\,A the dimension of the range of A . Let M\subset X be a non-empty set, n\in\mathbb{Z}_{+} . The Kolmogorov, linear and the Gelfand n -widths of M in X are defined, respectively, by the formulas d_{n}(M,\,X)=\inf_{L\in{\cal L}_{n}(X)}\sup_{x\in M}\inf_{y\in L}\|x-y\|, \lambda_{n}(M,\,X)=\inf_{A\in L(X,\,X),\,{\rm rk}\,A\leqslant n}\sup_{x\in M}\|x-Ax\|, d^{n}(M,\,X)=\inf_{x_{1}^{*},\,\dots,\,x_{n}^{*}\in X^{*}}\sup\Bigl\{\|x\|:\;x\in M\cap\Bigl(\cap_{j=1}^{n}\ker x_{j}^{*}\Bigr)\Bigr\} (in the last definition, we set \sup\varnothing=0 ). The problem of estimating the widhts of Sobolev classes on one-dimensional and multi-dimensional Lipschitz domains was studied in [ 21 , 22 , 23 , 24 , 25 , 26 , 27 , 28 , 29 , 30 , 31 , 34 , 32 , 33 , 35 ] (notice that the paper [ 35 ] is a recent result, which completely solves the Sobolev width problem on an interval; the critical case of W^{1}_{1} in L_{q} , 2<q<\infty , is considered). For John domains the order estimates of the widths were obtained in [ 36 ] and [ 37 ] ; these estimates are the same as for a cube. In [ 37 ] a more general case of weighted spaces was considered, and the proof was more complicated than in [ 36 ] . For r=1 , p=q and irregular domains (with zero angles) the estimates for linear widths were obtained by W.D. Evans, D.J. Harris and Y. Saito in [ 38 , 39 ] (more precisely, in these papers the approximation numbers were considered, but in the case of compact embeddings they are equal to the linear widths [ 40 ] ). In the case p<q , for a domain with a singular cusp of the form \{(x^{\prime},\,x_{d})\in\mathbb{R}^{d-1}\times\mathbb{R}:\,0<x_{d}<1,\,|x^{\prime}|<x_{d}^{\sigma}\} the order estimates for the Kolmogorov widths were obtained in [ 36 ] (here \sigma>1 , r+(\sigma(d-1)+1)(1/q-1/p)>0 ); the orders were found to be the same as for the cube. In [ 41 ] the domains \{(x^{\prime},\,x_{d})\in\mathbb{R}^{d-1}\times\mathbb{R}:\,0<x_{d}<1/2,\,|x^{\prime}|<x_{d}^{\sigma}|\log x_{d}|^{\alpha}\} were considered, where r=(\sigma(d-1)+1)(1/p-1/q) ; in this case, the orders may change already for small \alpha>0 . Now we give the definition of the entropy numbers. Let X , Y be normed spaces, and let T\in L(X,\,Y) , n\in\mathbb{N} . The entropy numbers of the operator T are defined by e_{n}(T)=\inf\Bigl\{\varepsilon>0:\;\exists y_{1},\,\dots,\,y_{2^{n-1}}\in Y:\;T(B_{X})\subset\cup_{i=1}^{2^{n-1}}(y_{i}+\varepsilon B_{Y})\Bigr\}. If \Omega is a cube, the order estimates for the entropy numbers of the embedding operator {\rm Id}:\tilde{\cal W}^{r}_{p}(\Omega)\to L_{q}(\Omega) were obtained in [ 42 ] , [ 43 ] . In [ 42 ] the proof depends on an approximation of a function f\in\tilde{\cal W}^{r}_{p}(\Omega) by a piecewise polynomial function with an appropriate partition. In [ 43 ] the discretization method and Schütt’s theorem [ 44 ] about estimates for the entropy numbers of finite-dimensional balls were applied. In [ 45 ] estimates for the entropy numbers were obtained for John domains; they are the same as for a cube (again, here the more general case of weighted spaces was considered; in the non-weighted case the proof can be significantly simplified). In [ 1 ] upper estimates for the entropy numbers were obtained in the case of Hölder domains and their anisotropic generalizations. Here these estimates will be improved. We also mention the paper [ 46 ] , where the upper estimates for the entropy and approximation numbers of the embedding operator of the anisotropic Sobolev space on a domain with a horn condition into the Orlicz space were obtained. Now we formilate the main results. Let the functions \varphi_{i}:(0,\,1]\rightarrow(0,\,1] (1\leqslant i\leqslant d-1) satisfy the following conditions: \displaystyle\varphi_{i}(t)\leqslant a_{*}t,\quad\varphi_{i}\uparrow,\quad\varphi_{i}(2t)\leqslant a_{*}\cdot\varphi_{i}(t),\quad 0\leqslant t\leqslant 1,\;1\leqslant i\leqslant d-1, (1) where a_{*}\geqslant 1 . Definition 1 . Let \Omega\subset\mathbb{R}^{d} be a domain. We write \Omega\in{\cal G}_{\varphi_{1},\,\dots,\,\varphi_{d-1}} if \displaystyle\Omega=\{(x^{\prime},\,x_{d}):\;x^{\prime}\in(0,\,1)^{d-1},\,0<x_{d}<\psi(x^{\prime})\}, (2) where \psi:[0,\,1]^{d-1}\rightarrow(0,\,\infty) ,