Upper bounds for the approximation of some classes of bivariate functions by triangular Fourier-Hermite sums in the space \(L_{2,p}(\mathbb{R}^2)\) (Q2288351)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Upper bounds for the approximation of some classes of bivariate functions by triangular Fourier-Hermite sums in the space \(L_{2,p}(\mathbb{R}^2)\)
scientific article

    Statements

    Upper bounds for the approximation of some classes of bivariate functions by triangular Fourier-Hermite sums in the space \(L_{2,p}(\mathbb{R}^2)\) (English)
    0 references
    17 January 2020
    0 references
    The authors study supprema of approximation of bivariate functions, generalizing research on the approximation of functions by algebraic polynomials on the real axis \(\mathbb{R}\) with Chebyshev weight \(\tilde{\rho}(x)=\exp{\{x^2\}}\). The main results are given in four theorems: \par Theorem 1. Let \(m,N\in\mathbb{N}\), \(r\in\mathbb{Z}_{+}\), \(0<p\leq 2\), \(h\in (0,1]\) and let \(q\) be a nonnegative measurable summable function on \((0,h)\) which does not vanish identically. Then \[\sup_{f\in L_{2,\rho}^{(r)}} \frac{N^rE_{N-1}(f)_{2,\rho}}{\left(\int_0^h \Omega_m^p(D^rf,t)_{2,\rho}q(t)dt\right)^{1/p}} = \frac{1}{\left\{\int_0^h (1-(1-t^2)^{N/2})^{mp}q(t)dt\right\}^{1/p}}. \] Theorem 2. Let \(m\in\mathbb{N}\), \(r\in\mathbb{Z}_{+}\). Then, for an arbitrary \(N\in\mathbb{N}\) \[\sup_{f\in L_{2,\rho}^{(r)}} \frac{N^r E_{N-1}(f)_{2,\rho}}{K(D^rf,\frac{1}{N^m})_{2,\rho}}=1.\] Theorem 3. Let \(m,N\in\mathbb{N}\), \(r\in\mathbb{Z}_{+}\), \(k=0,1,2,\ldots,N\), \(0<p\leq 2\), \(0<H<1\) and \(q\geq 0\) be a measurable function summable on \((0,H)\) which is not equivalent to zero. Then \[\gamma_{N(N+1)/2+k}(H_{2,p}^r (\Omega_m,q);L_{2,\rho}) = E_{N-1}(HW_{2,p}^r(\Omega_m,q))_{2,\rho}=\] \[=N^{-r}\left\{ \int_0^H (1-(1-t^2)^{N/2})^{mpp}q(t)dt\right\}^{-1/p},\] where \(\gamma(\cdot)\) is any of the following widths: Bernstein, Gelfand, Kolmogorov, linear and projectional. \par Theorem 4. Let \(\Phi\) be some majorant defining the class of functions \(W_{2,\rho}^r(K_m,\Phi)\), where \(r\in\mathbb{Z}_{+}\), \(m\in\mathbb{N}\). Then, for an arbitrary \(N\in\mathbb{N}\) and \(k=0,1,2,\ldots,N\) \[\gamma_{N(N+2)/2+k}(W_{2,\rho}^r(K;\Phi);L_{2,\rho}) = E_{N-1}(W_{2,\rho}^r(K;\Phi))_{2,\rho}=N^{-r}\Phi(N^{-m}),\] where \(\gamma_{nu}(\cdot)\) is any of the widths mentioned in Theorem 3. The specific definitions of the notations used would take to much space here.\par The main concept is the space \(L_{2,\rho}=L_{2,\rho}(\mathbb{R}^2)\); real squared summable functions on \(\mathbb{R}^2\) with weight \(\rho(x)=\exp{\{-(x^2+y^2)\}}\). The system \(\{H_k(x)H_l(x)\}_{k,l\in\mathbb{Z}_{+}}\) of Hermite polynomials is orthogonal on the entire plane \(\mathbb{R}^2\) with weight \(\rho\) and the double Fourier-Hermite series is given by \[f(x,y)=\sum_{k=0}^{\infty}\sum_{l=0}^{\infty}\,c_{kl}(f)H_k(x)h_l(y),\] with \[c_{kl}(f)=\int\int_{\mathbb{R}^2}\,\rho(x,y)f(x,y)H_k(x)H_l(y)dx dy\] convergence in \(L_{2,\rho}(\mathbb{R}^2)\)-sense.
    0 references
    0 references
    0 references
    0 references
    0 references
    upper bound
    0 references
    approximation of functions
    0 references
    triangular Fourier-Hermite sum
    0 references
    Jackson-Stechkin type inequality
    0 references
    \(K\)-functional
    0 references
    \(N\)-width
    0 references
    0 references