Solving underdetermined nonlinear equations by Newton-like method

Boris Polyak, Andrey Tremba

Published 2017-03-22Version 1

Newton method is one of the most powerful methods for finding solution of nonlinear equations. In its classical form it is applied for systems of $n$ equations with $n$ variables. However it can be modified for underdetermined equations (with $m<n$, $m$ being the number of equations). Theorems on solvability of such equations as well as conditions for convergence and rate of convergence of Newton-like methods are addressed in the paper. The results are applied to systems of quadratic equations, one-dimensional equations and inequalities.