Numerical Solution of a Class of Deficient Polynomial Systems

Abstract
Most systems of polynomials which arise in applications have fewer than the expected number of solutions. The amount of computation required to find all solutions of such a “deficient” system using current homotopy continuation methods is proportional to the expected number of solutions and, roughly, to the size of the system. Much time is wasted following paths which do not lead to solutions. We suggest methods for solving some deficient polynomial systems for which the amount of computational effort is instead proportional to the number of solutions.