Error formulas for ideal interpolation (Q2400086)

If \(\mathbb{F}\) is the field of real or complex numbers, denote by \(\mathbb{F}[x_1,x_2,\dots,x_n]\) the space of polynomials in \(n\) variables. An ideal interpolation is a linear idempotent projection \(P\) on \(\mathbb{F}[x_1,x_2,\dots,x_n]\), whose kernel is a polynomial ideal and which has a finite-rank. In the theory of ideal interpolation there exist several error formulas, i.e., representations of the difference \(f-Pf\). An important error formula was proposed by \textit{C. de Boor} [in: Approximation theory XI. Proceedings of the 11th international conference, Gatlinburg, TN, USA, May 18--22, 2004. Brentwood, TN: Nashboro Press. 59--91 (2005; Zbl 1126.41003)]. However this formula is not generally applicable, as it was shown by an counterexample by \textit{B. Shekhtman} [J. Approx. Theory 162, No. 7, 1398--1406 (2010; Zbl 1205.41005)]. In the present paper, an extension of C. de Boor's formula is introduced, named the ``normal'' formula. It is proved that this new formula is applicable in the case where the ideal has a lexicographic order reduced Gröbner basis. This new formula is applicable also for the example given by Shekhtman.