Comparing the degrees of unconstrained and shape preserving approximation by polynomials (Q320327): Difference between revisions

We denote by \(\mathbb{Y}_{s}\) the set of all collections \(Y_{s} = \{ y_{i} \}_{i=1}^{s},\) where \(y_{s+1} := -1 < y_{s} < \dots < y_{1} < 1 =: y_{0}\) and \( s \geq 1\) is given. By definition, \(f \in \Delta^{(1)}(Y_{s})\) if \( f \in C[-1,1]\) and \((-1)^{i}f\) is nondecreasing on \([y_{i+1}, y_{i}],\) \(i=0,1,\dots , s,\) and \(f \in \Delta^{(2)}(Y_{s})\) if \( f \in C[-1,1]\) and \((-1)^{i}f\) is convex on \([y_{i+1}, y_{i}],\) \(i=0,1,\dots , s,\) respectively. For \(f \in C[-1,1],\) we denote by \(E_{n}(f) := \inf_{p_{n} \in \mathbb{P}_{n}} \, \| f - p_{n} \|\) the degree of approximation of \(f\) by algebraic polynomials of degree \(< n\) (\(\mathbb{P}_{n}\) is the space of all algebraic polynomials of degree \(<n,\) and \(\| \cdot \|\) is the uniform norm on \(C[-1,1]\)). Further, for \(f \in \Delta^{(q)}(Y_{s}),\) \(q = 1, 2,\) we denote by \(E_{n}^{(q)}(f,Y_{s}) := \inf_{P_{n} \in \mathbb{P}_{n} \cap \Delta^{(q)}(Y_{s})} \, \| f - P_{n} \|\) the degree of best comonotone, respectively, coconvex approximation of \(f\) relative to \(Y_{s}.\) The authors prove the following results, which answer some former open questions. Theorem 1. If \(Y_{s} \in \mathbb{Y}_{s},\) \(s \geq 2,\) and \(1 < \alpha \leq 2,\) then there exist the constants \(c(\alpha, s)\) and \(N^{*}(\alpha, Y_{s})\) such that for the all functions \(f \in \Delta^{(1)}(Y_{s})\) we have \(n^{\alpha} E_{n}(f) \leq 1\) for \(n \geq s+2\) and \(n^{\alpha} E_{n}^{(1)}(f,Y_{s}) \leq c(\alpha, s)\) for \(n \geq N^{*}(\alpha, Y_{s}),\) respectively. Theorem 2. Given \(Y_{s} \in \mathbb{Y}_{s},\) \(s \geq 3,\) and \(2 < \alpha < 4.\) Then there exist the constants \(c(\alpha, s)\) and \(N^{*}(\alpha, Y_{s})\) such that the all functions \(f \in \Delta^{(2)}(Y_{s})\) satisfy that \(n^{\alpha} E_{n}(f) \leq 1\) for \(n \geq s+3\) and \(n^{\alpha} E_{n}^{(2)}(f,Y_{s}) \leq c(\alpha, s)\) for \(n \geq N^{*}(\alpha, Y_{s}).\)