Analysis of approximate factorization algorithm. I (Q1202761): Difference between revisions

The approximate factorization of a multivariate polynomial over the field \(\mathbb{C}\) of complex numbers is applied to polynomials with approximate coefficients such as floating-point numbers. In a paper by \textit{T. Sasaki}, \textit{M. Suzuki}, \textit{M. Kolář} and \textit{M. Sasaki} [Jap. J. Ind. Appl. Math. 8, No. 3, 357-375 (1991; Zbl 0757.12006)] two algorithms of approximate factorization have been proposed but not proved. The paper now gives a mathematical analysis of one of the algorithms presented provided that the numerical coefficients are calculated with sufficient accuracy. This algorithm determines the approximately irreducible factors by calculating approximately dependent linear combinations of roots of the form \(\lambda_ 1 \varphi_ 1^ i+ \cdots+ \lambda_ n \varphi_ n^ i\), where \(\varphi_ 1, \dots, \varphi_ n\) are the roots of the given polynomial with respect to one variable and \(i=1,2,\ldots\).