On the method of approximation by families of linear polynomial operators (Q5926109)

In the paper it is shown that the best approximation \(E_n(f)_p\) of \(f\in L^p\) \((0< p\leq+\infty)\) by trigonometric polynomials of order at most \(n\), is achieved in average by families of linear polynomial operators in the \(L^p\)-metric for all \(p\), \(0< p\leq+\infty\), e.g., there is proved, among others, that \[ \Biggl(\int^{2\pi}_0\|f- L_{n,\lambda} f\|^p_p d\lambda\Biggr)^{1/p}\leq C_p E_n(f)_p, \] where \[ L_{n,\lambda} f(x)= {1\over 2N+1} \sum^{2N}_{k=0} f\Biggl({2\pi k\over 2N+1}+ \lambda\Biggr) W_n \Biggl(x- {2\pi k\over 2N+1}- \lambda\Biggr), \] \(W_n\) is a kernel of de la Vallée-Poussin type, and \(N= [(1+ \delta)n]\) for some \(\delta\geq 0\) and positive integer \(n\). This is compared with approximation by Fourier means and interpolation means which is restricted to \(1\leq p\leq+\infty\) and \(p= +\infty\), respectively.