Average sampling in certain subspaces of Hilbert-Schmidt operators on L^2(\mathbb{R}^d)

Antonio G. García

arXiv: 2011.05871v2 · PDF · 22 January 2021
Categories: math.FA
Licence: CC BY 4.0 The site shows the text.

Abstract

The concept of translation of an operator allows to consider the analogous of shift-invariant subspaces in the class of Hilbert-Schmidt operators. Thus, we extend the concept of average sampling to this new setting, and we obtain the corresponding sampling formulas. The key point here is the use of the Weyl transform, a unitary mapping between the space of square integrable functions in the phase space \mathbb{R}^d\times \widehat{\mathbb{R}}^d and the Hilbert space of Hilbert-Schmidt operators on L^2(\mathbb{R}^d), which permits to take advantage of some well established sampling results.

Items 8

Definition

Definition 1

HCVB

A generalized stable sampling procedure in V_{S}^2 is a map \mathcal{S}_{\text{samp}}:V_{S}^2 \rightarrow \ell^2(\Lambda) defined as

T=\sum_{\lambda \in \Lambda} c(\lambda)\, \alpha_\lambda(S)\in V_{S}^2 \longmapsto \{s_T(\lambda)\}_{\lambda \in \Lambda} \text{ such that $s_T:= \mathbf{c} \ast_\Lambda \mathbf{q} \in \ell^2(\Lambda)$}\,,

where \mathbf{q} \in \ell^2(\Lambda) satisfies the conditions

0<\operatorname*{ess\,inf}_{\xi \in \widehat{\Lambda}} |\mathcal{F}_\sigma^\Lambda(\mathbf{q})(\xi)| \le \operatorname*{ess\,sup}_{\xi \in \widehat{\Lambda}} |\mathcal{F}_\sigma^\Lambda(\mathbf{q})(\xi)| <\infty\,.

Used by Proof

Theorem

Theorem 1

DW9A

Assume that a generalized stable sampling procedure \mathcal{S}_{\text{samp}} in V_{S}^2 is given as in Definition 1 with associated sequence \mathbf{q} \in \ell^2(\Lambda). Then, there exists a unique Hilbert-Schmidt operator H\in V_{S}^2 such that the sampling formula T=\sum_{\lambda \in \Lambda}s_T(\lambda)\, \alpha_\lambda(H) holds in V_{S}^2. The convergence of the series is unconditional in Hilbert-Schmidt norm.

Reciprocally, if a sampling formula like (?) holds in V_{S}^2 where s_T(\lambda)= \big(\mathbf{c} \ast_\Lambda \mathbf{q} \big)(\lambda), \lambda \in \Lambda, and \{\alpha_\lambda (H)\}_{\lambda \in \Lambda} is a Riesz basis for V_{S}^2, then the conditions in Eq.(1) are satisfied.

Proof of Theorem 1

The first part of the proof has been done above. For the second part, observe that \{\mathcal{T}_S^{-1}\big(\alpha_\lambda(H)\big)\}_{\lambda \in \Lambda} is a Riesz basis for \ell^2(\Lambda) with dual basis \{T_\lambda \mathbf{q}^*\}_{\lambda \in \Lambda} which implies conditions in (1).

Uses Definition 1

Definition

Definition 2

JM3V

A generalized stable sampling procedure in V_{\bf S}^2 is a map \mathcal{S}_{\text{samp}}:V_{\bf S}^2 \rightarrow \ell^2_{_M}(\Lambda) defined as

T=\sum_{m=1}^M\sum_{\lambda \in \Lambda} c(\lambda)\, \alpha_\lambda(S)\in V_{\bf S}^2 \longmapsto {\bf s_T}:=A \ast_\Lambda \mathbf{c} \in \ell^2_{_M}(\Lambda)\,,

where the matrix A=[a_{m,n}] \in \mathcal{M}_{_{M\times N}}\big(\ell^2(\Lambda)\big) satisfies the conditions:

0<\alpha_A:=\operatorname*{ess\,inf}_{\xi\in \widehat{\Lambda}} \lambda_{\text{min}}[\,\widehat{A}(\xi)^*\widehat{A}(\xi)] \le \beta_A:= \operatorname*{ess\,sup}_{\xi\in \widehat{\Lambda}} \lambda_{\text{max}}[\widehat{A}(\xi)^*\widehat{A}(\xi)]<+\infty\,.

Theorem

Theorem 2

MBCA

Assume that a generalized stable sampling procedure \mathcal{S}_{\text{samp}} with associated matrix A is given in V_{\bf S}^2 as in Definition 2. Then, there exist M\geq N elements H_m\in V_{\bf S}^2, m=1,2,\dots,M, such that the sampling formula

T=\sum_{m=1}^M\sum_{\lambda \in \Lambda} s_{_{T,m}}(\lambda)\, \alpha_\lambda(H_m) \quad \text{in $\mathcal{H}\mathcal{S}$-norm}

holds for each T\in V_{\bf S}^2 where \{\alpha_\lambda (H_m)\}_{\lambda \in \Lambda;\, m=1,2,\dots,M} is a frame for V_{\bf S}^2. The convergence of the series is unconditional in Hilbert-Schmidt norm.

Reciprocally, if a sampling formula like (2) holds in V_{\bf S}^2 where

{\bf s_T}(\lambda)=\big(s_{_{T,1}}(\lambda), s_{_{T,2}}(\lambda), \dots, s_{_{T,M}}(\lambda)\big)^\top:=\big(A\ast_\Lambda \mathbf{c}\big)(\lambda)\,, \quad \lambda \in \Lambda\,,

with \beta_A<+\infty, and \{\alpha_\lambda (H_m)\}_{\lambda \in \Lambda;\, m=1,2,\dots,M} is a frame for V_{\bf S}^2, then the left-hand condition in (2) also holds.

Proof of Theorem 2

The first part of the theorem has been proved above. Observe that the operators H_m\in V_{\bf S}^2, m=1,2,\dots,M, depend on the dual frames \big\{T_\lambda\,\mathbf{b}_{m}\big\}_{\lambda \in \Lambda;\, m=1,2,\dots,M} of the frame \big\{T_{\lambda}\,\mathbf{a}^*_{m}\big\}_{\lambda \in \Lambda;\, m=1,2,\dots,M}. Namely, H_m=L_{h_m} \in V_{\bf S}^2 with Weyl symbol h_m=\mathcal{T}_{a_{\bf S}}(\mathbf{b}_m) \in V_{a_{\bf S}}^2, m=1,2,\dots,M; \mathbf{b}_m denotes the m-th column of the matrix B with transfer matrix \widehat{B}. There are infinite dual frames whenever M>N; they are obtained from the left-inverses \widehat{B}(\xi) of \widehat{A}(\xi), i.e., \widehat{B}(\xi)\,\widehat{A}(\xi)=I_{_N}, a.e. \xi \in \widehat{\Lambda}, which are obtained, from the Moore-Penrose pseudo-inverse \widehat{A}(\xi)^\dag, by means of the N\times M matrices

\widehat{B}(\xi):=\widehat{A}(\xi)^\dag+C(\xi)\big[I_M-\widehat{A}(\xi)\widehat{A}(\xi)^\dag\big]\,, \text{\, \,a.e. $\xi \in \widehat{\Lambda}$}\,,

where C denotes any N\times M matrix with entries in L^\infty(\widehat{\Lambda}).

For the second part, we have that \{\mathcal{T}_{\bf S}^{-1}[\alpha_\lambda(H_m)]\}_{\lambda \in \Lambda;\, m=1,2,\dots,M} is a frame for \ell^2_{_N}(\Lambda) and

\mathbf{c}=\sum_{m=1}^M\sum_{\lambda \in \Lambda} s_{_{T,m}}(\lambda)\, \mathcal{T}_{\bf S}^{-1}[\alpha_\lambda(H_m)]=\sum_{m=1}^M\sum_{\lambda \in \Lambda} \,\big\langle \mathbf{c}, T_\lambda \mathbf{a}^*_{m} \big \rangle_{\ell^2_{_N}(\Lambda)} \mathcal{T}_{\bf S}^{-1}[\alpha_\lambda(H_m)]\quad \text{in $\ell^2_{_N}(\Lambda)$}\,.

Since \beta_A<+\infty, the sequence \big\{T_{\lambda}\,\mathbf{a}^*_{m}\big\}_{\lambda \in \Lambda;\, m=1,2,\dots,M} is a Bessel sequence for \ell^2_{_N}(\Lambda), and consequently (see Ref. [ole:16, Lemma 6.3.2]), a dual frame of \{\mathcal{T}_{\bf S}^{-1}[\alpha_\lambda(H_m)]\}_{\lambda \in \Lambda;\, m=1,2,\dots,M}; hence, \alpha_A>0.

Corollary

Corollary 3

P0BQ

In case M=N, the following statements are equivalent:

0<\operatorname*{ess\,inf}_{\xi\in \widehat{\Lambda}} \big|\det [\widehat{A}(\xi)]\big| \le \operatorname*{ess\,sup}_{\xi\in \widehat{\Lambda}} \big|\det [\widehat{A}(\xi)]\big|<+\infty
There exist N unique elements H_n, n=1,2,\dots, N, in V_{\bf S}^2 such that the associated sequence \big\{\alpha_\lambda(H_{n})\big\}_{\lambda\in \Lambda;\,n=1,2,\dots, N} is a Riesz basis for V_{\bf S}^2 and the sampling formula
T=\sum_{n=1}^N\sum_{\lambda \in \Lambda} s_{_{T,n}}(\lambda)\, \alpha_\lambda(H_n)\ \quad \text{ in $\mathcal{H}\mathcal{S}$}
holds for each T\in V_{\bf S}^2.

Moreover, the interpolation property s_{_{H_n,n'}}(\lambda)=\delta_{n,n'}\delta_{\lambda,0}, where \lambda \in \Lambda and n,n'=1,2,\dots,N, holds.

Proof of Corollary 3

In this case, the square matrix \widehat{A}(\xi) is invertible and statement 1. is equivalent to 0<\alpha_A \le \beta_A <+\infty; besides, any Riesz basis has a unique dual basis. The uniqueness of the coefficients in a Riesz basis expansion gives the interpolation property.

Source and licence

The text on this page is from arXiv:2011.05871v2 by Antonio G. García, under the licence CC BY 4.0. The corpus made these changes: it read the items out of the LaTeX source, it expanded the macros of the paper, it put each proof after its statement, and it converted the text to HTML. The formulas are the TeX of the paper, and KaTeX renders them. A formula or a reference can look different from the PDF. The page is published under CC BY 4.0.

The record of each item is in analysis/objects/2011/2011.05871.jsonl.