Local convergence for some high convergence order Newton-like methods with frozen derivatives (Q892992): Difference between revisions

The article deals with the local convergence analysis for the following Newton-like methods of approximate solving a nonlinear operator equation: \[ x_{n+1} = y_n - F'(x_n)^{-1}F(x_n), \quad y_n = \lambda(x_n), \] (Traub method), \[ \begin{aligned} y_n^{(1)}&= x_n - F'(x_n)^{-1}F(x_n), \\ y_n^{(2)} &= y_n^{(1)} - F'(x_n)^{-1}F(y_n^{(1)}), \\ &\;\vdots\\ y_n^{(m-1)} &= y_n^{(m-2)} - F'(x_n)^{-1}F(y_n^{(m-2)}), \\ x_{n+1} &= y_n^{(m-1)} - F'(x_n)^{-1}F(y_n^{(m-1)}), \end{aligned} \] (Amat-Busquier-Bermudez method), \[ \begin{aligned} y_n &= x_n - aF'(x_n)^{-1}F(x_n), \\ z_n^{(1)} &= x_n - \bigg[\dfrac1{2a} F'(y_n) + \bigg(1 - \frac1{2a}\bigg)F'(x_n)\bigg]^{-1}F(x_n), \\ z_n^{(2)} &= z_n^{(1)} - \varphi(a,x_n,y_n)F(z_n^{(1)}), \\ &\;\vdots \\ z_n^{(m)} &= z_n^{(m-1)} - \varphi(a,x_n,y_n)F(z_n^{(m-1)}), \\ x_{n+1} &= z_n^{(m)} - \varphi(a,x_n,y_n)F(z_n^{(m)}), \end{aligned} \] (Xiao-Yin method). Here \(\lambda(x)\) and \(\varphi(a,x,y)\) are some operators, for example, \(\lambda(x) = x - \alpha F'(x)^{-1}F(x)\) and \(\varphi(a,x,y) = 2\bigg[\dfrac1{2a}F'(y) + \bigg(1 - \dfrac1{2a}\bigg)F'(x)\bigg]^{-1} - F'(x)^{-1}\). The main results are three theorems about the local convergence of all these methods; the corresponding convergence conditions are formulated in terms of the first derivative of \(F\) and the Lipschitz constant for this derivative. Some numerical examples illustrate the results obtained.