Towards a higher degree \(F\)-transform (Q429366)

The notion of fuzzy transform (\(F\)-transform for short), introduced by \textit{I. Perfilieva} [Fuzzy Sets Syst. 157, No. 8, 993--1023 (2006; Zbl 1092.41022)], explains modeling with fuzzy IF-THEN rules as a specific transformation, thus bridging fuzzy modeling and the theory of linear (particularly, integral) transformations. The \(F\)-transform of a continuous function \(f:[a,b]\to \mathbb{R}\) is given by a vector \(F[f]=(F_1,\ldots, F_n)\), where \[ F_k=\frac{\int_a^b f(x)A_k(x)dx}{\int_a^b A_k(x)dx}\quad (k=1,\ldots,n) \] and \(A_1,\ldots,A_n\) are membership functions in a so-called fuzzy partition of \([a,b]\), defined by five specific axioms. The paper under review shows how the \(F\)-transform can be generalized to a \(F^{m}\)-transform, \(m\geq 0\), whose components are polynomials of degree \(m\). A detailed description of the \(F^{1}\)-transform, with linear polynomials as components, is given. Gaussian quadrature is applied in order to simplify the computation of these components. Finally, an inverse \(F^{m}\)-transform, \(m>0\), is introduced and shown to approximate the original function in several norms.