Variations on the Chebyshev and \(L^ q\) theories of best approximation (Q788939): Difference between revisions
From MaRDI portal
Set OpenAlex properties. |
ReferenceBot (talk | contribs) Changed an Item |
||
Property / cites work | |||
Property / cites work: Abstract K and J spaces / rank | |||
Normal rank | |||
Property / cites work | |||
Property / cites work: Restricted Range Approximation of <i>k</i>-Convex Functions in Monotone Norms / rank | |||
Normal rank |
Latest revision as of 10:53, 14 June 2024
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Variations on the Chebyshev and \(L^ q\) theories of best approximation |
scientific article |
Statements
Variations on the Chebyshev and \(L^ q\) theories of best approximation (English)
0 references
1982
0 references
For \(f\in C[0,1]\) and \(q\in [1,\infty)\) put \(\|| f\||_ q=\sup(\int^{b}_{a}| f(x)|^ q)^{1/q}\) where the supremum is taken over all intervals [a,b]\(\subseteq [0,1]\) for which \(f(x)>0\) on (a,b) or \(f(x)<0\) on (a,b). The functional \(\|| \cdot \||_{q^*}\) is defined similarly with the modification that the supremum is taken over all intervals [a,b]\(\subseteq [0,1]\), for which f(x)\(\geq 0\) on (a,b) or f(x)\(\leq 0\) on (a,b). These functionals, called gauges, are positively defined and absolutely homogeneous, but they not verify the triangle inequality. In this paper the authors restrict their attention to the case \(q=1\) (in this case the above defined functionals are denoted simply by \(\|| \cdot \||\) and \(\|| \cdot \||_*)\), although all the results (excepting Theorem 4.1) hold true for the case \(q\in [1,\infty)\). Concerning the relation of these gauges with the uniform topology on C[0,1] the authors prove that if \(f_ m\) and f are in C[0,1] and \(f_ m\) converges uniformly to f then \(\|| f\|| \leq \lim \inf \|| f_ m\|| \leq \lim \sup \|| f_ m\|| \leq \|| f\||_*\). The authors consider the problem of best approximation of continuous functions by polynomials of degree n with respect to the distance induced by these gauges. Although these gauges are not norms, one regains most of the geometric flavour of best uniform approximation in C[0,1]. Denote by \(\pi_ n\) the set of all polynomials of degree at most n and for \(f\in C[0,1]\) let \(d(f,\pi_ n)\) and \(d^*(f,\pi_ n)\) be the distance from f to \(\pi_ n\) with respect to \(\|| \cdot \||\) and \(\|| \cdot \||_*\), respectively. The infimum \(d(f,\pi_ n)\) is attained for all \(f\in C[0,1]\), but the infimum \(d^*(f,\pi_ n)\) may be unattained. The main result of the paper is the following Chebyshev type characterization of polynomials of best approximation: Theorem 3.1. For every \(f\in C[0,1]\) there exists a unique \(p\in \pi_ n\) such that \(\|| f-p\|| =d(f,\pi_ n)\) and this p is characterized by the following condition: there exist \(n+2\) nonempty open intervals \(I_ 1<I_ 2<...<I_{n+2}\) and \(\sigma\in \{-1,1\}\) fixed such that (a) \((-1)^ k\sigma(f-p)\geq 0\) on \(I_ k\) and (b) \((-1)^ k\sigma \int_{I_ k}(f-p)(x)\cdot dx\geq \|| f-p\||,\) for all \(k=1,...,n+2\). A similar result holds for \(\|| \cdot \||_*\). This condition is simpler to check than the well known conditions of characterization of best \(L^ q\)-approximation, namely, \[ \int^{1}_{0}| f(x)-p(x)|^{q-1}sgn(f(x)-p(x))x^ kdx=0, \] for \(k=0,1,...,n\).
0 references
best uniform approximation
0 references
gauges
0 references
uniform topology
0 references