Approximation, solution operators and quantale-valued metrics (Q2407702)

A generalized solution operator is a mapping abstractly describing a computational problem and its approximate solutions. It assigns a set of \(\varepsilon\)-approximations of a solution to the problem instance \(f\) and accuracy of approximation \(\varepsilon\). In this paper the author studies generalized solution operators for which the accuracy of approximation is described by elements of a complete lattice equipped with a compatible monoid structure, which is called a quantale. This study extends the results of \textit{A. G. Werschulz} [Aequationes Math. 26, 74--82 (1983; Zbl 0547.41029)], in which generalized solution operators defined over nonnegative reals were investigated, does provide with examples of computational problems for which the accuracy of approximation of a solution is measured by such objects, and shows that the sets of \(\varepsilon\)-approximations are, roughly, closed balls with radii \(\varepsilon\) with respect to a certain family of quantale-valued generalized metrics induced by a generalized solution operator. Some of the examples presented there seems to provide with a basis for future research on computational problems for which the accuracy of approximations of exact solutions is not described by nonnegative reals, but rather by some other partially ordered set. Distance spaces are also studied, interested results are pointed out.