Up to now we rely on singular to compute resultants of multivariate polynomials. There are faster ways.
Even computing the determinant of the sylvester matrix is usually (much) faster.
First i have implemented a trick for bivariate polynomials over the rationals (it could in principle work over any field with enough elements, but it is not clear that it is faster there). The trick consists on specialicing for several values of the surviving variable, compute the (univariate) resultant for them, and then reconstruct by lagrange interpolation.
It would also be worth to perform some benchmarks, and deduce a heuristic for the cases where the current method is beaten by the sylvester matrix determinant.
Up to now we rely on singular to compute resultants of multivariate polynomials. There are faster ways.
Even computing the determinant of the sylvester matrix is usually (much) faster.
First i have implemented a trick for bivariate polynomials over the rationals (it could in principle work over any field with enough elements, but it is not clear that it is faster there). The trick consists on specialicing for several values of the surviving variable, compute the (univariate) resultant for them, and then reconstruct by lagrange interpolation.
It would also be worth to perform some benchmarks, and deduce a heuristic for the cases where the current method is beaten by the sylvester matrix determinant.
CC: @saraedum
Component: algebra
Keywords: resultant, discriminant, polynomial, multivariate
Author: Miguel Marco
Branch/Commit: u/mmarco/ticket/16749 @
2ec35c2
Issue created by migration from https://trac.sagemath.org/ticket/16749