On discretization errors in nonlinear approximation problems (Q1819301)

Let \((X,\| \cdot X\|_ x)\) be a linear normed space of dim n and let \(A\subseteq X\) be a nonempty closed set. Let \(1\leq p\leq \infty\) and B be a compact set, in particular \(B=[c,d]\) if \(1\leq p\leq \infty\). Let C(B) be the space of all real-valued continuous functions on B, equipped with the \(L_ p\) norm \(\| f\|_ B=\max_{\xi \in B}| f(\xi)|\) if \(p=\infty\), \(\xi\in B\), \(\{\int^{d}_{c}| f(\xi)|^ pd\xi \}^{1/p}\) if \(f<1<p^{\infty}\) for \(f\in C(B)\). Let \(r\in C(B)\) be fixed and \(T: A\to C(B)\) be a continuous operator. The author considers the nonlinear approximation problem (P) Minimize \(\| r-Ta\|_ B\) on A with \(\rho =\inf \{\| r-Ta\|_ B,| a| \in A\},\) â\(\in A\) will be a solution of (P) if the infimum in (1) is achieved for â\(\in A\). Making two assumptions, the first one on a sequence of compact subsets of B and the next one on a sequence \(\{A_ k\}\) of closed subsets of \((X,\| \cdot \|_ x)\), the author proves in Theorem 1 a convergence problem, and in Theorem 2 he extends the results of E. W. Cheney for \(p=\infty\) and \(p=1\).