Bounded-degree factors of lacunary multivariate polynomials

Bruno Grenet (1)

1 ECO - Exact Computing

In this paper, we present a new method for computing bounded-degree factors of lacunary multivariate polynomials. In particular for polynomials over number fields, we give a new algorithm that takes as input a multivariate polynomial f in lacunary representation and a degree bound d and computes the irreducible factors of degree at most d of f in time polynomial in the lacunary size of f and in d. Our algorithm, which is valid for any field of zero characteristic, is based on a new gap theorem that enables reducing the problem to several instances of (a) the univariate case and (b) low-degree multivariate factorization. The reduction algorithms we propose are elementary in that they only manipulate the exponent vectors of the input polynomial. The proof of correctness and the complexity bounds rely on the Newton polytope of the polynomial, where the underlying valued field consists of Puiseux series in a single variable.

Domaines: Calcul formel [cs.SC]; Complexité [cs.CC]; Algorithme et structure de données [cs.DS]

Bruno Grenet. Bounded-degree factors of lacunary multivariate polynomials. Journal of Symbolic Computation, 2016, Special issue on the conference ISSAC 2014: Symbolic computation and computer algebra, 75, pp.171-192. ⟨10.1016/j.jsc.2015.11.013⟩. ⟨hal-01094727⟩