On the approximation of discrete functions by Fourier-Hahn sums (Q1180810)

Let \(N\geq 2\) be an integer, \(\mathbb{R}\) the set of real numbers. A mapping \(f: \{0,\dots,N-1\}\to\mathbb{R}\) is called a discrete function. Let \(\varphi_ 0(j),\dots,\varphi_{N-1}(j)\) be a finite sequence of functions forming an orthonormal system with respect to a weight \(\rho=\rho(j)>0\) on the set \(\{0,\dots,N-1\}\); \[ c_ k=c_ k(f):=\sum_{j=0}^{N-1}f(j)\varphi_ k(j)\rho(j) \qquad (0\leq k\leq N-1) \] are the Fourier coefficients of \(f\); \[ S_{n,N}(j)=S_{n,N}(f,j):=\sum _{k=0}^ n c_ k(f)\varphi_ k(j) \] are the partial sums of the Fourier series of \(f\) with respect to the system \(\{\varphi_ 0,\dots,\varphi_{N-1}\}\), where \(0\leq n\leq N-1\). The author raised the following problem: what is the maximum of the error on the set \(\{0,\dots,N-1\}\) if we substitue \(S_{n,N}(f,j)\) for \(f(j)\)? In other words, the author wants to estimate the maximal deviation \[ \max_{0\leq j\leq N-1}| f(j)-S_{n,N}(f,j)| \qquad (0\leq n\leq\psi(N)), \] where \(\psi(N)\) is an integral-valued function such that \(0\leq\psi(N)\leq N-1\). In the present paper the author solves this problem in the case when \(\{\varphi_ 0,\dots,\varphi_{N-1}\}\) consists of the Hahn orthogonal polynomials of a discrete variable.