On univariate cardinal interpolation by shifted splines (Q1263756): Difference between revisions

Let \(M_ n\) denote the centered univariate cardinal B-spline of order n. If \(\alpha\in {\mathbb{R}}\), then define \(M_{n,\alpha}(\cdot):=M_ n(\cdot +\alpha)\). Cardinal interpolation by \(M_{n,\alpha}\) is said to be correct if for each bounded sequence \((f(j))_{j\in {\mathbb{Z}}}\), there exists a unique bounded sequence \((a(j))_{j\in {\mathbb{Z}}}\) such that \(\sum_{j\in {\mathbb{Z}}}a(j)M_{n,\alpha}(k-j)=f(k)\) for all \(k\in {\mathbb{Z}}\). Fourier analytic methods are used to show that cardinal interpolation by \(M_{n,\alpha}\) is correct if and only if \(\alpha\neq mod {\mathbb{Z}}\). This provides an alternate proof of a result originally due to \textit{C. A. Micchelli} [Stud. Spline Funct. Approx. Theory, 203-250 (1976; Zbl 0343.41008)], and \textit{C. de Boor} and \textit{I. J. Schoenberg} [Lect. Notes Math. 501, 1-79 (1976; Zbl 0319.41010)]. Further, a convergence theorem of Schoenberg type [cf. \textit{I. J. Schoenberg}, Isr. J. Math. 16, 87-93 (1973; Zbl 0272.41005), and \textit{S. Riemenschneider}, ibid. 23, 339-346 (1976; Zbl 0332.41005)] is also established for shifted splines. The printed version of the paper contains several typographical errors, the most serious being the commencement of the proof of Theorem 3.1 (a) [p. 487]. The correct statement should read ``Since \(\{\) f(-j): \(j\in {\mathbb{Z}}\}\) are the Fourier coefficients of the measure \(\mu\), it follows from the weak convergence of its Fourier series that...''.