Designing optimal dual frames for ℓ p − \ell^{p}- average error optimization
Shankhadeep Mondal, Deguang Han, R. N. Mohapatra · 2025
Read Designing optimal dual frames for ℓ p − \ell^{p}- average error optimization by Shankhadeep Mondal, Deguang Han, R. N. Mohapatra (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-2508-07158.txt.
Introduction
Shankhadeep Mondal, Deguang Han, and Mohapatra ask which backup description of a signal best limits damage when parts of the signal are lost. Their paper appeared in two thousand twenty-five. Frames describe a signal with some slack, and dual frames rebuild it. When chunks vanish in transit, the choice of dual sets the error.
The team judges duals by the size of their error operators, under three different size measures. They seek the dual that cuts the average error over all loss patterns of a set size. For the gentlest measure they fully solve the problem for even, tight frames. Rules emerge for when the standard dual is the unique best pick. Links between the three measures are also drawn. The results serve speech, imaging, and any system that must rebuild from partial data.
Catalog record
| CREATOR | Shankhadeep Mondal, Deguang Han, R. N. Mohapatra |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 59 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2508-07158
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Designing optimal dual frames for ℓ p − \ell^{p}- average error optimization Shankhadeep Mondal, Deguang Han, R. N. Mohapatra Abstract In this paper, we investigates the problem of optimal dual frame selection for signal reconstruction in the presence of erasures. Unlike traditional approaches relying on left inverses, we evaluate performance through the norms of error operators, using the Frobenius norm, spectral radius, and numerical radius as measures. Our central focus is the characterization of dual frames that minimize the ℓ p − \ell^{p}- average under these error operator measurements over all possible erasure patterns. We provide conditions under which the canonical dual frame is uniquely optimal and extend our results to multiple erasures. In the Frobenius norm case, we offer a complete characterization for any number of erasures in uniform tight frames. The paper also examines interconnections between optimality criteria across different norm measures and gives sufficient conditions ensuring uniqueness of the optimal dual. 2010 MSC 42C15, 47B02, 94A12 1 Introduction The theory of frames has emerged as a powerful mathematical framework for robust signal representation, with wide-ranging applications in signal processing, wireless communications, and data transmission. One of the most compelling features of frames is their inherent redundancy, which provides resilience against the loss of information which is commonly referred to as erasures during transmission. When part of the encoded frame coefficients are lost due to adverse network conditions or channel imperfections, the redundant structure of frames allows for accurate reconstruction of the original signal, provided that an appropriate dual frame is chosen. A fundamental problem in this context is the selection of an optimal dual frame that minimizes the reconstruction error introduced by such erasures. Various norms of the associated error operators have been employed to quantify this reconstruction error [ 21 , 3 , 36 , 31 ] . Classical works have addressed the operator norm-based optimization of dual frames, notably for Parseval frames or for specific classes of erasures [ 30 ] . However, a comprehensive treatment of different operator-induced measurements in the broader setting of general frames remains relatively unexplored. In recent years, extensive research has been devoted to addressing erasure problems through the lens of frame theory. Casazza and Kovačević [ 24 ] explored equal norm tight frames in depth, analyzing their structure, construction, and robustness against erasures. Goyal, Kovačević, and Kelner [ 28 ] examined uniform tight frames from a coding-theoretic perspective, showing that they are optimal for handling single erasures. Holmes and Paulsen [ 30 ] introduced an operator norm-based framework to assess the optimality of dual frames, particularly focusing on Parseval frames and their canonical duals, and identified conditions under which the canonical dual minimizes the worst-case reconstruction error. This line of work was extended in [ 36 , 31 ] , where the uniqueness and topological properties of optimal duals were studied, especially under multiple erasure settings. Alternative approaches have also emerged. Bodmann and Paulsen [ 22 ] proposed an average-case measure for reconstruction error based on operator norms, while Pehlivan, Han, and Mohapatra [ 1 ] investigated the spectral radius as an error metric for one erasure, with further developments for two erasures in [ 21 , 27 ] . The numerical radius was analyzed in [ 2 ] , and an averaged combination of operator norm and numerical radius in [ 26 ] . While our work adopts a similar average-type measurement as in [ 22 ] , it significantly generalizes the setting by considering arbitrary frames and their duals, rather than restricting to Parseval frames and its canonical duals. Related investigations using probabilistic models and weight sequences were conducted in [ 34 , 35 ] , while broader studies on optimal dual pairs, including Parseval and general frames can be found in [ 32 , 33 , 17 ] . In recent years, several generalized frame structures have been developed, among which K- frames introduced by Găvruţa [ 5 ] have gained particular attention due to their role in reconstruction within \mathrm{Range}(K) . Structural and constructive results on K- frames and their duals appear in [ 6 , 7 , 8 , 9 , 10 , 11 , 12 , 13 , 14 , 15 ] , while erasure problems for K- frames are addressed in [ 44 , 7 ] . These studies lay the groundwork for further developments in designing optimal duals in the presence of erasure and reconstruction challenges. In this paper, we investigate the problem of optimal dual frame selection for a given finite frame, under erasure-induced error, using a measurement that is structurally different from the traditional min–max formulations in the literature. More precisely, we consider the \ell^{p} -average of the norm of the error operators over all possible m- element erasure patterns. This formulation offers a natural probabilistic interpretation and avoids the over-conservatism of worst-case models. For instance, rather than minimizing the worst-case error measure \max\bigg\{\mathcal{M}\big(T_{G}^{*}DT_{F}\big):D\in\mathcal{D}_{m}\bigg\}, where \mathcal{M} denotes a chosen measure of the error operator and \mathcal{D}_{m} is the collection of diagonal matrices with exactly m ones (indicating the retained coefficients) and N-m zeros on the main diagonal, as considered by Bodmann and Paulsen [ 22 ] , we instead focus on minimizing the \ell^{p} -average error , given by \left(\frac{1}{\binom{N}{m}}\sum_{D\in\mathcal{D}_{m}}\big[\mathcal{M}\big(T_{G}^{*}DT_{F}\big)\big]^{p}\right)^{\!\frac{1}{p}}, for p>1 . Where G is a dual frame of F and \mu is one of the error operator measures, namely the Frobenius norm, spectral radius, or numerical radius. Unlike previous studies that often restrict to Parseval frames and their canonical duals, our framework applies to general frames and arbitrary duals, significantly broadening the scope of the analysis. The structure of the paper is as follows: Section 2 introduces the necessary background on frames, duals, error operators, and matrix norms. Section 3 formalizes the average case measurement model and error formulation. Section 4 addresses the Frobenius norm and provides complete characterizations of optimal dual frames for any m- erasure scenario. Section 5 investigates the spectral radius as the error measure, and Section 6 focuses on the numerical radius. In Section 7, we study the relationships between optimal duals under the different norm measures and analyze their possible intersections and topological structure. This work contributes to the ongoing development of robust frame theory by extending optimal dual frame selection beyond canonical settings and introducing average-based, norm-sensitive optimization criteria with both theoretical and practical significance. 2 Preliminaries on erasure Throughout this article, we consider \mathcal{H}_{n} to be an n -dimensional Hilbert space. A finite collection of vectors F=\{f_{i}\}_{i=1}^{N} in \mathcal{H}_{n} is called a frame for \mathcal{H}_{n} if there exist constants A,B>0 such that A^{\prime}\|f\|^{2}\leq\sum_{i=1}^{N}|\langle f,f_{i}\rangle|^{2}\leq B^{\prime}\|f\|^{2},\hskip 10.00002pt\text{for all }f\in\mathcal{H}_{n}. The constants A^{\prime} and B^{\prime} are referred to as the lower and upper frame bounds , respectively. The optimal lower frame bound is the supremum of all possible lower bounds, and the optimal upper frame bound is the infimum over all upper bounds. A frame \{f_{i}\}_{i=1}^{N} is called tight if A^{\prime}=B^{\prime} , i.e., \sum_{i=1}^{N}|\langle f,f_{i}\rangle|^{2}=A^{\prime}\|f\|^{2},\hskip 10.00002pt\text{for all }f\in\mathcal{H}_{n}. In particular, if A^{\prime}=B^{\prime}=1 , then the frame is said to be a Parseval frame . Given a frame F for \mathcal{H}_{n}, the analysis operator T_{F}:\mathcal{H}_{n}\to\mathbb{C}^{N} is defined by T_{F}(f)=\{\langle f,f_{i}\rangle\}_{i=1}^{N}, and its adjoint T_{F}^{*}:\mathbb{C}^{N}\to\mathcal{H}_{n} , called the synthesis or pre-frame operator , is