Rate of convergence of Bernstein polynomials revisited

Approximation by polynomials (41A10) Rate of convergence, degree of approximation (41A25) Inverse theorems in approximation theory (41A27) Saturation in approximation theory (41A40)

Review text

Let \(C_{\alpha}\) be the collection of functions f, \(f\in C[0,1]\) for which \(\| (x(1-x))^{-\alpha /2}f(x)\|_{L_{\alpha}[0,1]}\equiv \| f\|_{C_{\alpha}}<\infty\) and \(C^ 2_{\alpha}\) a collection of functions in C[0,1] which are twice differentiable locally in (0,1) (f and f' are locally absolutely continuous) and for which the seminorm \(\| (x(1-x))^{1-(\alpha /2)}f(x)\|_{L_{\alpha}[0,1]}\equiv \| f\|_{C^ 2_{\alpha}}\) is finite. Furthermore, let the interpolation space \((C_{\alpha},C^ 2_{\alpha})_{\beta}\) be the collection of f for which \(K(t^ 2,f)/t^{\beta}<M(f),\) with norm \(\sup_{t}(K(t^ 2,f)/t^{\beta})\), where \(K(t^ 2,f)\) is given by \(K(t^ 2,f)=\inf_{f=f_ 1+f_ 2}(\| f_ 1\|_{C_{\alpha}}+t^ 2\| f_ 2\|_{C^ 2_{\alpha}}).\) For the Bernstein polynomials, given by \(B_ n(f,x)\equiv \sum^{n}_{k=0}f(k/n)\left( \begin{matrix} n\\ k\end{matrix} \right)x^ k(1-x)^{n-k},\) \(f\in C[0,1]\), \(X=x(1-x)\), \(0\leq \alpha <2\), \(0<\beta <2\) the author's main theorem now reads: The following conditions are equivalent: (a) \(X^{-\alpha /2}| B_ n(f,x)-f(x)| \leq Mn^{-\beta /2},\) (b) \(f\in (c_{\alpha},C^ 2_{\alpha})_{\beta}\), (c) \(X^{-\alpha /2}| f(x-t)-2f(x)+f(x+t)| \leq (t^ 2/X)^{\beta /2}.\) In view of known results, only the direction (c) implies (b) needs to be proved. For \(\beta =2\), the saturation case, the author shows the equivalence of (a) and (c).

Recommendations

Cites work

Cited in

(12)

This page was built for publication: Rate of convergence of Bernstein polynomials revisited

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1089540)