Quadratic vector equations (Q1940316)

The objective of this paper is centred on solving the equation (1) \(Mx = a + b(x,x)\), where \(a\) and the unknown \(x\) are componentwise nonnegative vectors, \(M\) is a nonsingular \(M\)-matrix, and \(b\) is a bilinear map from pairs of nonnegative vectors to nonnegative vectors. The main result in this respect is the following: Equation (1) has at least one solution if and only if there are a weakly positive map \(l:{{\mathcal R}^n} \to {{\mathcal R}^m}\) and a vector \(z \in {{\mathcal R}^m}\), \(z > 0\), such that for any \(x \geqslant 0\) it holds that \(l(x) \leqslant z\) implies \(l({M^{ - 1}}(a + b(x,x))) \leqslant z\); among its solutions, there is a minimal one. In a particular case, an algorithm derived for one equation of the class to another is adapted, which shows computational advantage with respect to the existing methods. Finally, some possible research lines, including the relationship among Newton-type methods and the cyclic reduction algorithm for unilateral quadratic equations are discussed.