The best multi-dimensional parametrization (Q5942186)

A system of \(n\) nonlinear algebraic or transcendental equations with \(m\) continuously varying parameters can be regarded as a system \(F_i(x_1,\ldots,x_n,p_1=x_{n+1},\ldots,p_m=x_{n+m})=0, i=1,\ldots,n\) in \(\mathbb{R}^{n+m}\). The set of all solutions of the system corresponding to various values for the parameters \(p\) is referred to as the surface of the set of solutions of the system. The method of continuation can be used to reduce the problem of finding a solution to this system to one of solving a system of linear algebraic equations whose matrix is an augmented Jacobi matrix of the system. The authors [Russ. Acad. Sci., Dokl., Math. 49, No. 1, 170-173 (1994); translation from Dokl. Akad. Nauk, Ross. Akad. Nauk 334, No. 5, 566-568 (1994; Zbl 0851.65027)] have proved that when \(m=1\), the best parametrization of extension of this system is the arclength of the solution set curve of the system, where by ``best'', the authors mean the parameter that makes the corresponding system of linear equations best conditioned. In this paper, they generalize this result to \(m>1\), proving that the system is best-posed if and only if the continuation parameters are chosen in the space tangent to the surface of the set of solutions.