Approximation of the \(r\)th differential operator by means of linear shape preserving operators of finite rank. (Q1421505)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Approximation of the \(r\)th differential operator by means of linear shape preserving operators of finite rank.
scientific article

    Statements

    Approximation of the \(r\)th differential operator by means of linear shape preserving operators of finite rank. (English)
    0 references
    26 January 2004
    0 references
    The author presents some results estimating the order of approximation of the \(r\)th derivative of a function by means of linear operators under different assumptions related to shape preserving properties. Let \(X= [0,1]\). Denote by \(\mathbb{C}^k(X)\), \(k\geq 0\), the space of all real-valued and \(k\)-times continuously differentiable functions on \(X\). Let \(D^i\) be the \(i\)th differential operator; \(\|\cdot\|\) denotes the sup-norm in \(\mathbb{C}(X)= \mathbb{C}^0(X)\), \(\| f\|= \sup_{x\in X}| f(x)|\). Denote by \(\Pi_k\) the subspace of \(\mathbb{C}(X)\) spanned by \(\{e_0,e_1,\dots, e_k\}\), where \(e_i(x)= x^i\). Let \(\sigma= (\sigma_i)_{i\geq 0}\) be a sequence with \(\sigma_i\in \{-1,0,1\}\), and let \(h\), \(k\) be two integers with \(0\leq h\leq k\) and \(\sigma_h\cdot\sigma_k\neq 0\). Let \(C_{h,k}(\sigma)= \{f\in \mathbb{C}^k(X): \sigma_i\cdot D^i f\geq 0\), \(i= h,\dots, k\}\). Let \(\Gamma= \{i: h\leq i< k\), \(\sigma_i\neq 0\), \(\sigma_{i+1}= 0\), \(\sigma_i\cdot\sigma_{i+2}\neq -1\}\). When \(\Gamma=\phi\), then \(C_{h,k}(\sigma)\) is called a cone of type I. When \(\Gamma\neq\phi\), then we call \(C_{h,k}(\sigma)\) a cone of type II. A linear operator mapping \(\mathbb{C}(X)\) into a linear space of finite dimension \(n+1\) is called an operator of finite rank \(n+1\). The main result is the following Theorem: Let \(C_{h,k}(\sigma)\) be a cone of type II and \(r\in\Gamma\). Let \(L_n: \mathbb{C}^k(X)\to \mathbb{C}^r(X)\) be a linear operator of finite rank \(n+1\), such that (1) \(L_n(C_{h,k}(\sigma))\subset C_{r,n}(\sigma)\); (2) \(D^rL_n e_r= D^r e_r\); (3) if \(r> 0\), then \(L_n(\Pi_{r-1})\subset \Pi_{r-1}\). Then \(\sum^k_{p=r+1} {1\over p!} \| D^r L_n e_p- D^r e_p\|\geq {1\over \tau(n+1)^{k-r}}\), where the constant \(\tau> 1\) does not depend on \(n\). Finally, an example of an operator with the best order of approximation is given.
    0 references
    0 references
    Korovkin-type results
    0 references
    Shape preserving property
    0 references
    Order of approximation
    0 references
    0 references

    Identifiers