Painless Construction of Unconditional Bases for Anisotropic Modulation and Triebel-Lizorkin Type Spaces
Morten Nielsen · 2025
Read Painless Construction of Unconditional Bases for Anisotropic Modulation and Triebel-Lizorkin Type Spaces by Morten Nielsen (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-2503-16995.txt.
Introduction
Nielsen builds ready-to-use wave bases for function spaces that treat directions unequally, with every dial set in advance. Such bases split a signal into smooth, localized pieces and underpin compression and analysis. His 2025 construction tensors together one-dimensional brushlet functions, themselves built from local trigonometric pieces in the frequency domain.
The resulting system is orthonormal, not redundant, and provably serves the whole family of Triebel-Lizorkin and Besov spaces, including their alpha-tuned variants. The second half studies best short approximations with these bases, proving both achievability and limit results. The inverse limits lean on the system's non-redundancy. The author calls the recipe painless because nothing is left to tune.
Catalog record
| CREATOR | Morten Nielsen |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 106 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-16995
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Painless Construction of Unconditional Bases for Anisotropic Modulation and Triebel-Lizorkin Type Spaces Morten Nielsen Abstract . We construct smooth localized orthonormal bases compatible with anisotropic Triebel-Lizorkin and Besov type spaces on ℝ d \mathbb{R}^{d} . The construction is based on tensor products of so-called univariate brushlet functions that are based on local trigonometric bases in the frequency domain, and the construction is painless in the sense that all parameters for the construction are explicitly specified. It is shown that the associated decomposition system form unconditional bases for the full family of Triebel–Lizorkin and Besov type spaces, including for the so-called α \alpha -modulation and α \alpha -Triebel-Lizorkin spaces. In the second part of the paper we study nonlinear m m -term approximation with the constructed bases, where direct Jackson and Bernstein inequalities for m m -term approximation with the tensor brushlet system in α \alpha -modulation and α \alpha -Triebel-Lizorkin spaces are derived. The inverse Bernstein estimates rely heavily on the fact that the constructed system is non-redundant. 1991 Mathematics Subject Classification 42B35, 42C15, 42C40 1. Introduction Well-localized unconditional bases for function spaces defined on \mathbb{R}^{d} play a cental role for many applications as the bases offer a framework for discretisation of a wide variety of continuous problems relevant for, e.g., numerical analysis and mathematical modeling. One classical example is the discretisation of Calderón–Zygmund operators using a smooth wavelet orthonormal basis, see [ 38 ] . A well-localized basis often also supports a simple characterization of smoothness in terms of certain sparseness conditions relative to expansions in the basis. For example, smoothness measured on the Besov scale can be shown to corresponds to a certain sparseness condition on a corresponding orthonormal wavelet expansion [ 37 ] . Moreover, such norm characterizations often also allow us to make a natural connection to m -term nonlinear approximation using the basis by identifying certain smoothness spaces as nonlinear approximation spaces [ 24 , 25 , 33 , 32 ] . As a consequence we may gain specific knowledge on how to compress smooth functions using sparse representations of the functions in the unconditional basis [ 16 , 17 ] . From the point of view of mathematical modeling, it is desirable to have as flexible tools as possible as it allows one to incorporate more refined structure of various real-world phenomena in the model. Function spaces in anisotropic and mixed-norm settings have therefore attached considerable interest recently, see for example [ 29 , 13 , 4 , 10 ] and references therein. However, when it comes to construction of unconditional bases, a serious design challenge is to make the basis compatible with a desired anisotropic structure. One illustrative such example is the challenge of designing orthonormal wavelet bases for an arbitrary dilation matrix, see, e.g., [ 9 , 45 ] . An addition design challenge comes into play when one also desires that the constructed basis should be able to capture smoothness relative to some predefined notion of smoothness. Often this constraint translates to a specific requirement on the time-frequency structure of the basis. For example, Besov spaces are constructed using a dyadic decomposition of the frequency space, while the family of modulation spaces introduced by Feichtinger [ 21 ] correspond to a uniform decomposition of the frequency space. So for characterization of Besov spaces one needs a wavelet-type basis with a dyadic frequency structure in order to characterize smoothness, while for modulation spaces one needs a Gabor type system compatible with a uniform decomposition of the frequency space. In the present paper we address some of the mentioned challenges by proposing a general procedure for construction of a large and flexible family of tensor product orthonormal bases for L_{2}(\mathbb{R}^{d}) that can be adapted to be compatible with an arbitrary anisotropy on \mathbb{R}^{d} . We will show that the construction can easily be adapted to provide unconditional bases for a large and diverse family of smoothness spaces constructed by the so-called decomposition method. The theory of decomposition spaces [ 20 , 22 ] provides a very general framework for the study of function spaces that can be adapted to almost any anisotropic setting, and the special case of decomposition spaces in the frequency domain [ 8 ] , provide a very flexible model for smoothness that allows for a unified study of classical smoothness spaces such as modulation spaces, (aniotropic) Besov spaces, and (aniotropic) Triebel-Lizorkin spaces. In this paper we focus on the family of \alpha -modulation and \alpha -Triebel–Lizorkin spaces. These two families of smoothness spaces depend on a parameter \alpha\in[0,1] that governs the geometric time-frequency structure of the spaces. For \alpha=0 , we obtain the modulation spaces, while for \alpha=1 we obtain anisotropic Besov and Triebel-Lizorkin spaces, respectively. For \alpha\in(0,1) , we obtain spaces corresponding to a certain “polynomial” type decomposition of the time-frequency space. The reader may consult [ 8 , 1 , 23 , 15 , 5 , 12 ] for further discussion on \alpha -modulation type spaces. The main contribution of the present paper is to offer an easy explicit construction of a family of universal tensor product orthonormal bases for L_{2}(\mathbb{R}^{d}) , based on so-called univariate brushlet systems. The univariate brushlets are incredibly flexible orthonormal function systems based on local Fourier bases as introduced by Coifman and Meyer [ 14 ] , and by Malvar [ 35 ] for applications in signal processing. These systems were further developed by Wickerhauser [ 46 ] . Laeng pointed out in [ 34 ] that it is possible to map a local Fourier basis by the Fourier transform to a new basis with compact support in the frequency domain. In [ 36 ] , Coifman and F. G. Meyer studied similar systems, coining the term bruslets, using the bases introduced by Wickerhauser. Let us give an overview of the contributions from our study. (1) The construction of the tensor brushlet orthonormal bases is “painless” in the sense that all parameters for the construction are completely specified for any given anisotropy, \alpha -paramter, and dimension d\geq 2 . The construction is presented in Section 3 . (2) The construction provides the first example of an orthonormal system extending to universal unconditional bases for the full scale of anisotropic \alpha -modulation and \alpha -Triebel-Lizorkin type spaces on \mathbb{R}^{d} , regardless of the dimension d\geq 2 . (3) The basis supports a full characterization of the norm in anisotropic \alpha -modulation and \alpha -Triebel-Lizorkin spaces. The characterizations are proved in Section 4 . (4) The non-redundancy of the basis implies that inverse Bernstein estimates for nonlinear m -term approximation with the brushlet system can be obtained with approximation error measured in anisotropic \alpha -modulation and \alpha -Triebel-Lizorkin spaces. The approximation results are presented in Section 5 . Notation Throughout the article, positive constants, denoted by c , c^{\prime} or similar, may vary at every occurrence. For constants where a dependence on parameters q_{1},q_{2},\ldots,q_{K}, is essential for the context, it will be stated as c:=c(q_{1},q_{2},\ldots,q_{K}) . We will use the notation A\asymp B to indicate that the quantities A and B are equivalent in the sense that there exists constants c,c^{\prime}\in(0,\infty) such that cA\leq B\leq c^{\prime}A . 2. The anisotropic setup and preliminaries Let