Broyden's method in Hilbert space

From MaRDI portal





\textit{C. G. Broyden} [Math. Comput. 19, 577-593 (1965; Zbl 0131.139)] introduced an iteration quasi-Newton method to solve a finite-dimensional system of nonlinear equations, without the computation of the derivatives at each iteration, but by a good approximation of the derivatives with rank-one operators. Let X and Y be Hilbert spaces and let \(F: X\to Y\) be a nonlinear mapping. The author studies the local convergence behaviour of Broyden's method for the solution of the equation: \(F(x)=0_ Y\), with the iteration scheme: \(x_{i+1}=x_ i-B_ i^{-1}F(x_ i)\), where \(B_ i\in B(X,Y)\), the space of all bounded operators defined on X with the range in Y. The author proves that the algorithm yields a linear rate of convergence and under additional assumptions, it even yields a superlinear convergence rate. It is clear that these quasi-Newton algorithms shall have a day in the matter of solving equations, and the method proved here, is very actual.




Cited in
(30)








This page was built for publication: Broyden's method in Hilbert space

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3734179)