\(K_a\)-convergence and Korovkin type approximation (Q1669689)

The author is interested in Korovkin-type theorems in approximation theory (the famous `first' theorem of this kind is usually referred to as `the Bohman-Korovkin theorem'). This theorem ensures that convergence of sequences of positive linear operators on the space of continuous real functions \(C(I)\), defined on a compact subset \(I\) of the real numbers, for the `basis' functions \[ e_0(x)=1,\;e_1(x)=x,\;e_2(x)=x^2, \] implies convergence for all functions \(f\in C(I)\). This theorem does not only hold for ordinary convergence, but also for so-called statistical convergence (for the definition, see the paper, page 108). The author now turns to \(K_a\)-convergence (introduced by \textit{M. Lazić} and \textit{V. Jovović} [Publ. Elektroteh. Fak., Univ. Beogr., Ser. Mat. 4, 81--87 (1993; Zbl 0808.40002)]). Let \(a=(a_k)\) and \(x=(x_k)\) be number sequences and define \(y=(y_k)=K_a(x)\), the \(K_a\)-transform of \(x\), by \[ y_k=\sum_{i=1}^k\,a_{k-i+1}x_i\quad (k=1,2,3,\ldots). \] The sequence \(x\) is then said to be \(K_a\)-convergent to \(\ell\) if its \(K_a\)-transform \(y\) has limit \(\ell\), i.e., \(\lim_k\,y_k=\ell\). The main result is then: Theorem 2.4. Let \(a=(a_k)\) be a number sequence and \(\sum_{k=1}^{\infty}\,| a_k| <\infty\). Suppose that \((L_i)\) is a sequence of positive linear operators acting from \(C(I)\) into itself, satisfying \[ \lim_k\left\| \sum_{i=1}^k\, a_{k-i+1}L_i(e_r)-e_r\right\| =0\quad (r=0,1,2). \] Then, for all \(f\in C(I)\), we have \[ K_a-\lim_k\| L_k(f)-f\| =0,\,\hbox{i.e.}, \] \[ \lim_k\left\| \sum_{i=1}^k\, a_{k-i+1}L_i(f)-f\right\| =0. \] The layout of the paper is as follows: \begin{itemize} \item[--] Introduction: Definitions and background material concerning the three types of convergence that appear. \item[--] A Korovkin-type approximation: Formulation and proof of the main result. Moreover, an example of a sequence \((a_k)\) is given for which Theorem 2.4 holds, but not the Korovkin theorem for ordinary convergence, statistical convergence and convergence using weighted sections of the form \(\frac{1}{k}\sum_{i=n}^{n+k-1}\,L_i(f)\), where \(n\rightarrow\infty\). \item[--] Rate of convergence: A result concerning the rate of \(K_a\)-convergence of \(L_k(f)\) to \(f\), using the usual modulus of continuity of \(f\). \end{itemize}