Third-order iterative methods without using any Fréchet derivative. (Q1408391): Difference between revisions

Many general problems can be often reduced to finding zeros of nonlinear equations. These roots cannot, in general, be expressed in closed form. One of the most important techniques for solving such equations are usage of iterative processes starting from some pivot initial approximation and computing successive approaches \(x^{i+1} := F(x^i)\). The Newton method is an example of such a method of second order of convergence. Let \(F : B \subset X \to X\) be a nonlinear operator defined on a convex subset \(B\) of the Banach space \(X\). The classical Chebyshev (\(b=0\)), Halley (\(b=1/2\)) and SuperHalley (\(b=1\)) methods can be defined as \[ x^{n+1} = x^n - (I + 1/2\;LF(x^n)(I - b\;LF)(x^n))^{-1} (F'(x^n))^{-1} F(x^n), \] where \(LF\) can be computed by the first and second Fréchet derivatives of \(F\). In the paper a modification of these classical third order iterative schemes is given. The main advantage of the developed methods is that they do not need to evaluate any derivatives of \(F\) and have the same range of convergence as the mentioned classical methods have.