A Haar-type theory of best uniform approximation with constraints (Q1187332)

The literature of approximation theory contains numerous results concerning best approximation under a variety of constraints. These results mainly concentrate on investigating problems in \(L_ \infty\)- norm, see, e.g., papers by \textit{B. L. Chalmers} [\((*)\) Trans. Amer. Math. Soc. 166, 309-316 (1972; Zbl 0252.41005)] and \textit{B. L. Chalmers} and \textit{G. D. Taylor} [Jahresber. Dtsch. Math.-Ver. 81, 49-86 (1979; Zbl 0438.41001)]. Since in most cases existence can be easily established the truly interesting question consists in studying the uniqueness of best approximants. In \((*)\) Chalmers introduced a general method of investigating uniqueness of best approximations with constraints, which provided a unified approach to the problem. However, this approach essentially provided only sufficient conditions for uniqueness of best constrained approximation. In this paper we are concerned with developing a Haar-type theory for constrained \(L_ \infty\)-approximation, that gives necessary and sufficient conditions for uniqueness. In this sense our paper is closer in spirit to a recent paper by \textit{A. Pinkus} and \textit{H. Strauss} [IMA J. Numer. Anal. 8, No. 1, 1-22 (1988; Zbl 0662.41019)], where necessary and sufficient conditions for uniqueness were given in the special case of best \(L_ \infty\)-approximation with coefficient constraints. Our goal is to provide similar characterizations of uniqueness in the general setting of linear constraints. In order to impose the constraints we use linear operators, instead of linear functionals used in \((*)\). This approach, while preserving generality of constraints, will provide us with a convenient tool leading to technically simple characteristic of uniqueness.