On the rate of approximation by generalized de la Vallée Poussin type matrix transform means of Walsh-Fourier series (Q6107469)

Let \(\mathbb{N}_+\) be the set of natural numbers and denote \(\mathbb{N}=\mathbb{N}_+\cup\{0\}\). Let us consider \(\mu\) the normalized Haar measure on \(\mathbb{Z}_2\) given by \(\mu(\{0\})=\mu(\{1\})=1/2\) and the Walsh group \(G=\{x=(x_1, \dots, x_k ,\dots)/x_k\in\{0,1\}, k\in\mathbb{N}\}\). Let us consider the Lebesgue spaces on \(G\), \(L^p(G)\), with the corresponding norm \(\| \; \|_p\), where \(1\leq p\leq \infty\) and let us introduce the modulus of continuity defined by \(\omega_p(f,\delta)=\sup_{|t|<\delta}\|f(\;\dot+t)-f(\;)\|_p\), if \(\delta>0\) for \(f\in L^p(G)\) where \(|x|=\sum_{i=0}^\infty \frac{x_i}{2^{i+1}},\) \(\forall x\in G\). Then, the Lipschitz classes are defined as \[ \mathrm{Lip}(\alpha,p,G)=\{f\in L^p(G)\;:\omega_p(f,\delta)=O(\delta^\alpha), \text{as} \; \delta\to 0\}. \] The Rademacher functions are defined as \(r_k(x)=(-1)^{x_k}\), \(x\in G\), \(k\in\mathbb{N}\). Observing that every natural number \(n\) can uniquely be expressed as \(n=\sum_{i=0}^\infty n_i2^i\), \(n_i\in \{0,1\}\), \(i\in\mathbb{N}\), where only a finite numbers of \(n_i's\) are different from zero, let us introduce the order of \(n\) given by \(|n|=\max\{j\in\mathbb{N}\;:n_j\not=0\}\). Then, the Walsh-Paley system \(\{w_n\}_{n\geq 0}\) is defined as \[ w_n(x)=\begin{cases} 1, &\text{if } n=0,\\ \prod_{k=0}^{\infty}r_k^{n_k}(x)=(-1)^{\sum_{k=0}^{|n|}n_kx_k},& \text{if } n> 0 \end{cases} \] and as in the classical case Fourier-Walsh coefficients \(\hat{f}(k)\), \(n\)th partial sums of Fourier series \(S_n(f)\), \(n\)th Dirichlet kernels \(D_n\), and \(n\)th Fejer kernels \(K_n\) are defined as \[ \hat{f}(k)=\int_Gf(x)w_n(x)d\mu(x),\\ \ S_n(f)=\sum_{k=0}^n\hat{f}(k)w_k,\\ \ D_n=\sum_{k=0}^{n-1}w_k \] and \(K_n=\frac{1}{n}\sum_{k=1}^nD_k\). Then, if \(T=(t_{k,n})_{k,n=1}^\infty\) is an infinite matrix of nonnegative numbers, such that, \(t_{k,n}=0\), for \(k>n\), the \((m,n)\) matrix transform de la Vallé Poussin is defined by \[ \sigma^{T}_{m,n}(f)=\sum_{k=m}^nt_{k,n}S_k(f). \] In this context the authors obtain estimates difference between \(\sigma^{T}_{m,n}(f)\) and the corresponding function \(f\). More precisely. Theorem. Let \(f\in L^p(G)\) and \(m,n\in\mathbb{N}_+\), where \(m<n\). Let the finite sequences \(\{t_{k,n}: m\leq k\leq n\}\) of nonegative numbers be nondecreasing for all \(n\). If \(\sum_{k=m}^nt_{k,n}=1\) and \(t_{n,n}=O(n^{-1})\), then \(\|\sigma^{T}_{m,n}(f)-f \|_1\leq c\omega_1(f,2^{-|m|})\). Similar results with variant hypotheses about the finite sequences \(\{t_{k,n}: m\leq k\leq n\}\) are obtained if \(f\in C(G)\), or \(f\in L^p(G)\), where \(1<p<\infty\). Finally, estimates for Lipschitz functions are summarized in the following statement. Theorem. Let \(f\in Lip(\alpha,p,G)\) where \(1<p<\infty\), for some \(\alpha>0\). Suppose that \(\sum_{k=m}^nt_{k,n}=1\). \begin{itemize} \item[{a)}] If the finite sequence \(\{t_{k,n}: m\leq k\leq n\}\) is nondecreasing \(\forall n\), supposing boundedness of \(nt_{n,n}\) and \(t_{n,n}=O(n^{-1})\), \item[{b)}] If the finite sequence \(\{t_{k,n}: m\leq k\leq n\}\) is nonincreasing \(\forall n\), supposing boundedness of \(mt_{m,n}\) and \(t_{m,n}=O(m^{-1})\), \end{itemize} then, in both cases, \(\|\sigma^{T}_{m,n}(f)-f \|_p=O(m^{-\alpha})\).