A Kantorovich-type analysis of Broyden's method using recurrent functions (Q2380857): Difference between revisions

The paper is concerned with the problem of approximating iteratively a locally unique solution of a nonlinear equation in Banach spaces by \textit{C. G. Broyden}'s iterative method [Math. Comput. 19, 577--593 (1965; Zbl 0131.13905), see also \textit{D. M. Hwang} and \textit{C. T. Kelly}, SIAM J. Optim. 2, No.~3, 505--532 (1992; Zbl 0760.65057)]. Broyden's method includes, as a particular case, the well-known Newton-Kantorovich method (if the inverse of Jacobian in Newton's method is replaced by a bounded linear operator). In contrast to earlier works on this topic, in Banach space setting, the authors use a more accurate so called L-center-Lipschitz condition, instead the stronger K-Lipschitz condition. This provides more precise majorizing sequences, whose convergence requires weaker sufficient convergence conditions than before. The idea of using the recurrent functions and a weaker assumption than the one given before using Newton-Kantorovich hypothesis lead to new convergence results obtained by the authors. The semilocal convergence of Broyden's method is analysed and the radius of convergence for this method is improved. Applications are given for solving three types of equations: a polynomial equation, a nonlinear Fedholm integral equation and a nonlinear boundary value problem for a second order differential equation.