Best approximation by the inverse of a monotone polynomial and the location problem (Q1604403)

Let \(\mathcal L = \{f\in C[0,1]:f\) is non-decreasing, \(f(0) = 1\) and \(f(1) = 1\}\). One considers the family of polynomials \( M = \{ p : p\in \mathcal P_n\), \(s_i p^{(l_i)} \geq 0\) for \(i=1,2,\dots ,\lambda\); \(p(0) =0\), \(p(1) = 1 \}\) where \(1<l_1<l_2<\dots <l_\lambda \leq n\), \(s_1 =1\) and \(s_i =1\) or \(s_i=-1\) are given. The author studies the problem of best approximation of a function \(f\in \mathcal L\) by elements in the set \(M^{-1}= \{p^{-1} : p\in M\}\) in the uniform norm and in the \(L_1\) norm. He proves: If \(f\in C[0,1]\), \(f(0) = \) and \(f(1) =1\) and if \(p^{-1}\in M^{-1}\) is the element of local uniform best approximation of \(f\) then \(p^{-1}\) is the unique element of global best uniform approximation by elements in \(M^{-1}\). A similar result holds for the approximation in the norm \(L_1\) of an element \(f\in \mathcal L\) by elements in \(M ^{-1}\). The author shows also the connection of his results with the location problem studied by \textit{V. I. Berdyshev} starting with 1993 [see, e.g., Russ. Acad. Sci., Dokl, Math. 46, No. 1, 157-160 (1993; Zbl 0804.41004) or Russ. J. Numer. Anal. Math. Math. Model. 8, No. 2, 83-100 (1993; Zbl 0818.65007)].