On positive and copositive polynomial and spline approximation in \(L_ p[-1,1],\quad 0<p<\infty\) (Q1925091): Difference between revisions

Suppose that \(f\in L_p[-1,1]\), \(0<p<\infty\), changes its sign finitely many times in the interval and we wish to approximate it by polynomials and splines which change sign at exactly the same points where \(f\) does. This type of approximation is called copositive approximation and it is the subject of the paper. The authors show that the degree of copositive approximation by polynomials of degree not exceeding \(n\), can be estimated by \(C\omega_\varphi(f,n^{-1})_p\), where the latter is the Ditzian-Totik modulus of continuity in the \(L_p\) norm. This is best possible because it is known that one cannot achieve the degree \(\omega_2(f,n^{-1})_p\) as was shown by \textit{S. P. Zhou} [Approximation Theory Appl. 9, No. 2, 104-110 (1993; Zbl 0773.41018)] for \(1<p<\infty\) and for \(f\) with at least one change of sign in the interval; and is shown in this paper for all \(0<p<\infty\) and even when \(f\) is positive throughout. This is in big contrast to the approximation of a nonnegative \(f\in C[-1,1]\) by nonnegative polynomials, in the uniform norm, where one can easily show that one can achieve the degree of unconstrained approximation. Similar best possible estimates are given for splines with \(n\) equidistant knots.