A simple test qualifying the accuracy of Horner's rule for polynomials
The authors consider the problem of evaluating a polynomial with floating point arithmetic. An evaluation scheme is defined to be faithful if it returns either the rounded up or the rounded down value of the exact result. Faithful rounding is shown to be more accurate than bounding the number of units in the last place of the error. The main results of the paper propose tight sufficient conditions to guarantee faithful rounding on one step of Horner's scheme. The conditions are validated with the IEEE standard for floating point arithmetic using the Coq automatic proof checker. Three programs in Maple, Java and C are also presented that compute an upper bound and an absolute error bound on an arbitrary polynomial and an arbitrary domain for the indeterminate. Subject to a certain criterion, the programs guarantee faithful evaluation, as illustrated by two examples.
- Precise evaluation of a polynomial at a point given in staggered correction format
- Horner's rule for interval evaluation revisited
- Transformation of polynomial evaluation to a pipeline via Horner's rule
- Accelerations of Zhao's methods for the numerical inversion of Laplace transform
- scientific article; zbMATH DE number 1206081 (Why is no real title available?)
- Accurate Horner methods in real and complex floating-point arithmetic
This page was built for publication: A simple test qualifying the accuracy of Horner's rule for polynomials
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2387741)
