An improved convergence theorem for a class of secant-like methods (Q2381342): Difference between revisions

Secant-like methods in Banach spaces are studied for solving the nonlinear operator equation \(F(x)= 0\), where \(F\) is defined on an open convex domain \(\Omega\) of a Banach space \(X\) with values in a Banach space \(Y\). Secant-like methods are described by the following algorithm \[ \begin{gathered} x_{-1},\;x_0\in\Omega\text{ pre-chosen},\\ y_n= (1-\lambda) x_n+ (1-\lambda) x_{n-1},\quad \lambda\in [0,1],\\ x_{n+1}= x_n- [y_n, x_n; F]^{-1} F(x_n),\end{gathered}\tag{1} \] where \([y, x; F]\) denotes the divided difference of order one for the operator \(F\) at the points \(x\) and \(y\) \((x\neq y)\). Since \([y_n,x_n;F]= F'(x_n)\) with \(y_n= x_n\), (1) is reduced to the secant method if \(\lambda= 0\) and to the Newton method if \(\lambda= 1\). The convergence of the secant method is usually proved by means of majorizing sequences. Assuming that the first-order divided difference is Hölder continuous, new natural and weaker conditions to ensure the convergence area proposed and a new recurrence relation are established as well as an improved semilocal convergence theorem for a class of secant-like methods is proved. Finally, two simple examples confirming the obtained theoretical results are presented.