Computational accuracy in determining the sign of an algebraic quantity

Abstract

AbstractThis paper considers the computational accuracy when performing a numerical computation of an algebraic quantity to determine the sign. For an algebraic function α({di}) of data {di} we introduce the (n, m, l) degree to discuss a method for its estimation. It is shown that when data {di} are given in fixed decimal point form with finite figures, we can correctly determine the sign of the value of α with approximately m times more figures.