Rate of convergence of higher-order methods (Q1946151)

Considering a quadratic expansion of \(F(x_k + s)\) around the \(k\)-th iterate \(x_k\), the authors derive a unified framework for the Halley class of methods and Schröder's method for solving systems of nonlinear equations \(F(x) = 0\). These methods are single point iterative methods using the first and second derivatives, but with third and second order of convergence rate, respectively. The methods in the Halley class require solutions of two linear systems of equations for each iteration. To improve their efficiency, the authors use the unified framework to derive inexact methods that solve the first system exactly and the second approximately through a few linear fixed point iterations. The authors prove the rates of convergence for these exact and inexact methods with results illustrated through some numerical experiments.