An alternating projection algorithm for the “approximate” GCD calculation

Abstract

In the paper an approach is proposed for calculating the “best” approximate GCD of a set of coprime polynomials. The algorithm is motivated by the factorisation of the Sylvester resultant matrix of polynomial sets with nontrivial GCD. In the (generic) case of coprime polynomial sets considered here the aim is to minimise the norm of the residual error matrix of the inexact factorisation in order to compute the “optimal” approximate GCD. A least-squares alternating projection algorithm is proposed as an alternative to the solution of the corresponding optimisation problem via nonlinear programming techniques. The special structure of the problem in this case, however, means that the algorithm can be reduced to a sequence of standard subspace projections and hence no need arises to compute gradient vectors, Hessian matrices or optimal step-lengths. An estimate of the asymptotic convergence rate of the algorithm is finally established via the inclination of two subspaces.