\(u\)-generation: solving systems of polynomials equation-by-equation (Q6140899)

In this article, the authors develop a new one-stage procedure dubbed \textit{u-generation} for a step in the equation-by-equation cascade. The core is a geometric construction that relies on a homotopy in \(P^{n+1}\), thus introducing one new variable \(u\). The new method improves the efficiency of equation-by-equation homotopy continuation methods for solving polynomial systems. These improvements may be applied to the basic algorithms of numerical algebraic geometry in the settings of both projective and multiprojective varieties. In Section 3, the authors extend the approach, albeit in a nontrivial way, to multihomogeneous systems: homotopies are transplanted from \(P^{n_1}\times \cdots \times P^{n_k}\) to \(P^{n_1+1}\times \cdots \times P^{n_k+1}\). The computational experiments demonstrate significant savings obtained on several benchmark systems. The authors also present an extended case study on maximum likelihood estimation for rank-constrained symmetric \(n\times n\) matrices, in which multiprojective u-generation allows us to complete the list of ML degrees for \(n = 6\).